framework,version,device,op_name,kernel_source,phase,batch_size,seq_len,num_tokens,d_model,d_conv,num_k_heads,head_k_dim,num_v_heads,head_v_dim,model_name,latency
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1,1,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.0072063997387886046
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1,1,1,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.004527999833226204
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1,1,1,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.0048096001148223875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,2,1,2,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.003958399966359138
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,2,1,2,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.005062400177121162
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,4,1,4,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.004025600105524063
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2,2,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.007625599950551986
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4,4,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.006963200122117996
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8,8,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.005910399928689003
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16,16,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.005849599838256836
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32,32,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.005875200033187866
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,64,64,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.005846399813890457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,128,128,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.005824000015854836
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,256,256,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.006307200342416763
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,512,512,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.00599679984152317
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1024,1024,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.00605119988322258
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2048,2048,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.007900799810886382
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4096,4096,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.011187200248241425
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8192,8192,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.023097600042819976
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16384,16384,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.05960639715194702
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32768,32768,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.12656320333480836
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1,2,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.011231999844312668
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2,4,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.011235199868679047
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4,8,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.011075200140476226
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8,16,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.008371199667453765
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16,32,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.008390399813652038
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32,64,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.008483199775218964
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,64,128,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.008495999872684479
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,128,256,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.0084927998483181
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,256,512,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.008336000144481659
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,512,1024,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.008403199911117553
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1024,2048,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.008636800199747085
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2048,4096,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.01154559999704361
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,4,1,4,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.00642239972949028
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4096,8192,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.024383999407291412
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8192,16384,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.05581759810447693
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16384,32768,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.11741440296173096
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,8,1,8,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.003990399837493897
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32768,65536,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.22956159114837646
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1,4,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.01881600022315979
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2,8,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.018697600066661834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4,16,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.018534399569034576
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8,32,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.013420799374580383
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16,64,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.013382400572299957
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32,128,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.013462400436401368
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,64,256,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.013475200533866883
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,128,512,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.01345600038766861
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,256,1024,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.013596799969673157
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,512,2048,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.013596799969673157
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1024,4096,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.013900800049304963
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2048,8192,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.027055999636650084
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4096,16384,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.0576960027217865
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8192,32768,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.11891839504241944
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16384,65536,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.22983040809631347
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32768,131072,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.4536032199859619
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1,8,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.03450239896774292
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2,16,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.034857600927352905
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4,32,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.03375039994716644
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8,64,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.02361920028924942
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16,128,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.02364480048418045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32,256,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.023651200532913207
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,8,1,8,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.007046400010585785
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,16,1,16,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.004553600028157234
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,64,512,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.023622399568557738
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,128,1024,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.023632000386714935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,256,2048,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.023500800132751465
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,512,4096,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.023766399919986726
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1024,8192,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.031439998745918275
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2048,16384,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.060115200281143186
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4096,32768,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.12292159795761108
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8192,65536,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.22922239303588868
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16384,131072,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.4463871955871582
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32768,262144,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.884819221496582
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1,16,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.06562880277633668
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2,32,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.06515520215034484
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4,64,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.06432639956474304
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8,128,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.043849599361419675
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16,256,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.04382719993591309
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,16,1,16,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.009679999947547913
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,32,512,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.04392319917678833
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,64,1024,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.04385280013084412
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,128,2048,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.04379520118236542
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,256,4096,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.043971198797225955
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,32,1,32,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.004198399931192398
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,512,8192,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.04789760112762451
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1024,16384,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.0676479995250702
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2048,32768,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.1317088007926941
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4096,65536,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.2423072099685669
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8192,131072,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.45433602333068845
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16384,262144,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.8833951950073242
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1,32,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.12620799541473388
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2,64,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.1261407971382141
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,32,1,32,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.025088000297546386
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4,128,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.12401920557022095
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8,256,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.08526719808578491
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,64,1,64,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.004854400083422661
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,16,512,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.08529279828071594
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,32,1024,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.08533440232276916
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,64,2048,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.0852735996246338
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,128,4096,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.0852512001991272
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,256,8192,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.08727359771728516
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,512,16384,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.09613440036773682
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1024,32768,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.14232959747314453
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2048,65536,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.260918402671814
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4096,131072,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.4849247932434082
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8192,262144,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.9063648223876953
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1,64,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.2507200002670288
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,64,1,64,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.046492800116539
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2,128,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.25030720233917236
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,128,1,128,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.006172800064086914
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4,256,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.24748480319976807
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,8,512,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.16720000505447388
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,16,1024,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.16708799600601196
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,32,2048,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.16713600158691405
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,64,4096,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.16712960004806518
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,128,8192,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.1690559983253479
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,256,16384,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.17135039567947388
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,512,32768,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.19082560539245605
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1024,65536,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.2756639957427979
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2048,131072,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.5166368007659912
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4096,262144,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.9551648139953614
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,128,1,128,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.08916479945182801
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,256,1,256,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.008607999980449676
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1,1,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.00732479989528656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2,2,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.007280000299215317
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4,4,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.007257600128650665
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8,8,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.006003199890255928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16,16,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.006041600182652473
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32,32,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.005984000116586685
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,64,64,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.005974400043487549
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,128,128,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.0061760000884532925
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,256,256,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.006041600182652473
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,512,512,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.00615679994225502
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1024,1024,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.006153599917888641
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2048,2048,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.00790719985961914
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4096,4096,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.01114879995584488
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8192,8192,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.023695999383926393
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16384,16384,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.05915840268135071
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32768,32768,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.12683199644088744
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1,2,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.01120320037007332
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2,4,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.01114879995584488
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4,8,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.011023999750614166
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8,16,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.008448000252246856
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16,32,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.008390399813652038
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32,64,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.008377599716186523
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,64,128,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.008495999872684479
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,128,256,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.008601599931716919
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,256,512,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.008566399663686752
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,512,1024,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.008627200126647949
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1024,2048,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.00873280018568039
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2048,4096,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.011750400066375732
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4096,8192,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.02446399927139282
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8192,16384,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.05634880065917969
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16384,32768,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.11758079528808593
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32768,65536,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.22980160713195802
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,256,1,256,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.17411839962005615
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1,4,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.01892479956150055
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2,8,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.018892799317836762
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4,16,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.01860159933567047
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8,32,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.013468800485134125
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,512,1,512,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.01353919953107834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16,64,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.013631999492645264
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32,128,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.01353600025177002
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,64,256,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.01353600025177002
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,128,512,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.013439999520778656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,256,1024,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.013603200018405915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,512,2048,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.013622400164604188
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1024,4096,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.013926400244235993
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2048,8192,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.0268095999956131
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4096,16384,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.05737919807434082
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8192,32768,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.1189087986946106
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16384,65536,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.22962241172790526
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32768,131072,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.45548157691955565
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1,8,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.03472320139408112
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2,16,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.03447360098361969
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4,32,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.03402880132198334
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8,64,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.023868800699710847
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16,128,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.023907199501991272
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32,256,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.02383359968662262
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,64,512,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.023827199637889863
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,128,1024,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.023820799589157105
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,256,2048,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.02375359982252121
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,512,4096,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.023977600038051605
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1024,8192,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.03131200075149536
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2048,16384,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.06099200248718262
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4096,32768,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.12345919609069825
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8192,65536,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.22893760204315186
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,512,1,512,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.342960000038147
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16384,131072,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.4465439796447754
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1024,1,1024,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.02338239997625351
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32768,262144,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.885536003112793
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1,16,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.06572480201721191
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2,32,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.06522560119628906
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4,64,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.0643775999546051
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8,128,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.04413439929485321
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16,256,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.04413439929485321
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,32,512,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.04411199986934662
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,64,1024,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.04410560131072998
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,128,2048,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.044095999002456664
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,256,4096,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.04409280121326446
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,512,8192,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.04795520007610321
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1024,16384,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.06749759912490845
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2048,32768,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.1318336009979248
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4096,65536,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.24167358875274658
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8192,131072,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.4549568176269531
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1024,1,1024,1024,4,16,128,16,128,Qwen/Qwen3.5-0.8B,0.6831456184387207
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16384,262144,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.8828831672668457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1,1,1,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.004073600098490715
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1,1,1,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.005033599957823753
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1,32,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.1271391987800598
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,2,1,2,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.003977600112557411
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,2,1,2,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.005011200159788132
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2,64,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.12702720165252684
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4,128,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.12549439668655396
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,4,1,4,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.0038816001266241073
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,4,1,4,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.005606399849057198
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8,256,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.08510720133781433
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,16,512,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.08532480001449586
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,8,1,8,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.004383999854326248
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,8,1,8,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.006883200258016586
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,32,1024,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.085343998670578
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,16,1,16,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.004057599976658821
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,16,1,16,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.00950080007314682
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,64,2048,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.08521919846534728
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,128,4096,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.08535040020942689
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,32,1,32,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.004003199934959412
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,32,1,32,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.025043201446533204
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,256,8192,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.0874176025390625
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,512,16384,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.09649919867515563
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,64,1,64,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.004732799902558327
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,64,1,64,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.046803200244903566
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1024,32768,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.1276927947998047
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2048,65536,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.25999999046325684
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,128,1,128,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.005913599953055381
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,128,1,128,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.08864960074424744
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,256,1,256,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.008486399799585343
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,256,1,256,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.17373440265655518
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4096,131072,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.483900785446167
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,512,1,512,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.013439999520778656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,512,1,512,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.3438463926315308
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1024,1,1024,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.023369599878787995
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8192,262144,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.9076736450195313
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1024,1,1024,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.6850207805633545
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1,1,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.007417599856853485
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1,64,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.2507008075714111
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2,128,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.25039680004119874
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4,256,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.24741439819335936
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,8,512,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.16723519563674927
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,16,1024,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.1671839952468872
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,32,2048,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.1673151969909668
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,64,4096,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.16725759506225585
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,128,8192,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.1693279981613159
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,256,16384,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.17184000015258788
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,512,32768,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.19107520580291748
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1024,65536,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.274780797958374
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2048,131072,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.5149504184722901
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4096,262144,2048,4,16,128,16,128,Qwen/Qwen3.5-2B,0.9528896331787109
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1,1,1,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005020799860358238
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2,2,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.007372800260782242
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4,4,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.007231999933719635
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8,8,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005977600067853928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16,16,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005977600067853928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32,32,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.006006399914622307
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,64,64,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.00599679984152317
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,128,128,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.006041600182652473
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,256,256,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005990400165319443
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,512,512,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005977600067853928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1024,1024,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.006137600168585777
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2048,2048,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.00799039974808693
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4096,4096,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.01112319976091385
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8192,8192,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.023212799429893495
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16384,16384,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.05903040170669556
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32768,32768,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.1264672040939331
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1,2,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.011043199896812439
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2,4,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.011132799834012986
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4,8,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.011023999750614166
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8,16,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.008259200304746629
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16,32,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.008326400071382523
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32,64,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.008323200047016144
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,64,128,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.008345600217580795
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,128,256,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.008428800106048583
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,256,512,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.008320000022649765
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,512,1024,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.00846719965338707
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1024,2048,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.008617600053548813
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2048,4096,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.01162559986114502
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4096,8192,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.02436479926109314
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1,1,1,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.0051807999610900875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8192,16384,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.05602560043334961
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16384,32768,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.1178879976272583
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,2,1,2,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005862399935722351
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32768,65536,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.22965118885040284
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1,4,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.01892160028219223
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2,8,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.018908800184726716
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4,16,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.018598400056362152
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8,32,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.013558399677276612
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16,64,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.013603200018405915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32,128,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.013612799346446991
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,64,256,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.013558399677276612
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,128,512,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.013574400544166565
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,256,1024,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.01353919953107834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,512,2048,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.013638399541378021
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1024,4096,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.013859200477600097
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2048,8192,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.026800000667572023
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4096,16384,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.05713919997215271
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8192,32768,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.11889599561691284
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16384,65536,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.22980160713195802
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32768,131072,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.4534048080444336
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1,8,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.03475840091705322
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2,16,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.03421440124511719
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4,32,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.03377279937267304
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8,64,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.023686400055885314
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,2,1,2,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005603199824690819
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16,128,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.023628799617290495
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32,256,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.023632000386714935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,64,512,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.023583999276161192
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,128,1024,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.0235615998506546
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,256,2048,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.023577600717544556
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,4,1,4,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005209600180387497
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,512,4096,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.023785600066185
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1024,8192,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.031404799222946166
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2048,16384,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.060659199953079224
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4096,32768,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.11439039707183837
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8192,65536,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.22929599285125732
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16384,131072,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.4463871955871582
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1,16,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.0654591977596283
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2,32,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.06504319906234741
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4,64,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.06418880224227905
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8,128,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.044073599576950076
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16,256,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.04407039880752563
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,32,512,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.04406720101833343
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,64,1024,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.043977600336074826
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,128,2048,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.043968001008033754
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,256,4096,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.044073599576950076
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,512,8192,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.04806399941444397
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1024,16384,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.06716799736022949
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2048,32768,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.1318079948425293
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,4,1,4,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.007180800288915634
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4096,65536,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.24209918975830078
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,8,1,8,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.004915200173854828
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8192,131072,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.452623987197876
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1,32,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.12632639408111573
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2,64,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.1264415979385376
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4,128,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.12488640546798706
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8,256,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.0847711980342865
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,16,512,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.08475840091705322
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,32,1024,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.08472639918327332
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,64,2048,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.08467199802398681
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,128,4096,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.0847711980342865
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,256,8192,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.08701120018959045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,512,16384,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.09627839922904968
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1024,32768,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.14363839626312255
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,8,1,8,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.009769599884748459
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2048,65536,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.2595168113708496
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,16,1,16,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005078399926424027
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4096,131072,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.47912960052490233
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1,64,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.2492703914642334
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2,128,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.24881279468536377
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4,256,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.24574720859527588
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,8,512,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.16611520051956177
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,16,1,16,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.025411200523376466
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,16,1024,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.16601920127868652
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,32,2048,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.16613119840621948
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,64,4096,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.16607359647750855
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,32,1,32,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.0051968000829219815
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,128,8192,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.16823359727859497
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,256,16384,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.17041280269622802
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,512,32768,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.1913696050643921
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1024,65536,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.2752896070480347
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2048,131072,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.5116672039031982
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,32,1,32,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.04615679979324341
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1,1,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.007465600222349167
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,64,1,64,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.005187200009822845
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2,2,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.00740479975938797
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4,4,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.007187200337648391
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8,8,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005990400165319443
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16,16,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005958399921655655
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32,32,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005964799970388413
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,64,64,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005999999865889549
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,128,128,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.006169600039720535
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,256,256,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.006070400029420853
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,512,512,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.006089600175619126
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1024,1024,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.0063008002936840056
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2048,2048,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.007971200346946716
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4096,4096,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.011273600161075592
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8192,8192,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.023446400463581086
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16384,16384,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.05956799983978271
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32768,32768,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.12699520587921143
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1,2,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.01141119971871376
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2,4,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.011334399878978729
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4,8,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.01117440015077591
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8,16,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.008601599931716919
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16,32,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.008582399785518646
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32,64,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.008556800335645676
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,64,128,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.008470399677753449
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,128,256,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.00852160006761551
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,256,512,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.008451200276613235
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,512,1024,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.008540800213813782
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1024,2048,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.008723200112581254
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2048,4096,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.011673600226640702
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4096,8192,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.02431039959192276
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8192,16384,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.05623679757118225
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16384,32768,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.11802560091018677
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,64,1,64,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.08923199772834778
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32768,65536,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.23023359775543212
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1,4,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.018812799453735353
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2,8,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.018889600038528444
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,128,1,128,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.006060799956321717
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4,16,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.018643200397491455
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8,32,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.013449600338935852
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16,64,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.013488000631332398
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32,128,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.01356479972600937
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,64,256,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.013631999492645264
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,128,512,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.01361600011587143
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,256,1024,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.013596799969673157
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,512,2048,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.013699199259281158
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1024,4096,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.013993600010871887
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2048,8192,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.02717440128326416
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4096,16384,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.057532799243927
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8192,32768,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.11887999773025512
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16384,65536,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.22990720272064208
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32768,131072,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.4533696174621582
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1,8,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.034748798608779906
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2,16,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.034329599142074584
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4,32,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.033913600444793704
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8,64,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.024278399348258973
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16,128,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.023945599794387817
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32,256,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.023868800699710847
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,64,512,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.023708799481391908
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,128,1024,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.02364480048418045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,256,2048,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.023708799481391908
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,512,4096,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.023875199258327484
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1024,8192,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.031116798520088196
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,128,1,128,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.17310080528259278
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2048,16384,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.060399997234344485
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4096,32768,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.12381119728088379
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,256,1,256,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.008502399921417237
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8192,65536,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.22946560382843018
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16384,131072,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.4469024181365967
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1,16,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.06558719873428345
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2,32,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.06518399715423584
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4,64,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.06376000046730042
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8,128,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.043884798884391785
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16,256,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.043952000141143796
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,32,512,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.04389120042324066
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,64,1024,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.043907201290130614
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,128,2048,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.044012799859046936
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,256,4096,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.044147199392318724
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,512,8192,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.047830399870872495
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1024,16384,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.06809599995613098
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2048,32768,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.13234879970550537
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4096,65536,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.242195200920105
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,256,1,256,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.34275519847869873
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8192,131072,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.4539231777191162
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,512,1,512,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.013574400544166565
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1,32,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.12656320333480836
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2,64,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.12643519639968873
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4,128,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.12501120567321777
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8,256,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.08489279747009278
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,16,512,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.08491520285606384
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,32,1024,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.08478720188140869
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,64,2048,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.08473280072212219
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,128,4096,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.08469759821891784
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,256,8192,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.08672639727592468
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,512,16384,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.09644160270690919
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1024,32768,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.14239039421081542
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2048,65536,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.2603967905044556
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,512,1,512,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.6799680233001709
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4096,131072,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.4819488048553467
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1024,1,1024,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,0.02327360063791275
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1,64,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.2489408016204834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2,128,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.24871039390563965
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4,256,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.24618880748748778
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,8,512,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.16638400554656982
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,16,1024,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.16628799438476563
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,32,2048,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.16599999666213988
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1024,1,1024,2560,4,16,128,32,128,Qwen/Qwen3.5-4B,1.3610848426818847
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,64,4096,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.16604479551315307
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,128,8192,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.1680608034133911
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1,1,1,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.00517439991235733
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1,1,1,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005049600079655647
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,2,1,2,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005049600079655647
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,2,1,2,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005587200075387955
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,256,16384,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.17020479440689087
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,512,32768,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.1866752028465271
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,4,1,4,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005075199902057648
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,4,1,4,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.0068960003554821014
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,8,1,8,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005177599936723709
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,8,1,8,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.009536000341176987
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1024,65536,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.2753439903259277
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,16,1,16,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005036799982190132
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,16,1,16,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.02476799935102463
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,32,1,32,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005305600166320801
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,32,1,32,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.04628799855709076
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2048,131072,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.5093152046203613
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,64,1,64,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.005212799832224846
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,64,1,64,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.08872320055961609
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,128,1,128,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.006041600182652473
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,128,1,128,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.17343039512634278
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,256,1,256,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.00854720026254654
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,256,1,256,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.34250240325927733
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,512,1,512,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.013471999764442444
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,512,1,512,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.6797311782836915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1024,1,1024,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,0.023366400599479677
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1,1,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.007427199929952622
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1024,1,1024,4096,4,16,128,32,128,Qwen/Qwen3.5-9B,1.3600095748901366
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1,1,1,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.00517439991235733
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1,1,1,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.005833600088953972
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,2,1,2,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.005244800075888634
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2,2,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.007196799665689468
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,2,1,2,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.006204799935221672
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4,4,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.007078400254249573
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,4,1,4,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.005311999842524529
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8,8,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.00589120015501976
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16,16,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.00570559985935688
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32,32,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.006006399914622307
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,64,64,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.005980800092220307
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,128,128,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.006041600182652473
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,256,256,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.005984000116586685
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,512,512,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.006073600053787232
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1024,1024,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.00626240000128746
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2048,2048,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.007929600030183791
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4096,4096,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.011289600282907486
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8192,8192,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.02364480048418045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16384,16384,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.06001600027084351
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32768,32768,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.126528000831604
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,4,1,4,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.00814720019698143
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1,2,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.011171200126409531
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,8,1,8,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.005471999943256378
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2,4,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.011126399785280228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4,8,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.011030399799346923
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8,16,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.008451200276613235
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16,32,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.008377599716186523
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32,64,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.008377599716186523
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,64,128,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.00843520015478134
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,128,256,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.00835840031504631
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,256,512,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.008399999886751174
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,512,1024,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.008489599823951722
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1024,2048,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.008643200248479843
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2048,4096,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.011711999773979187
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4096,8192,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.024073599278926848
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8192,16384,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.05611519813537598
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16384,32768,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.11799039840698242
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32768,65536,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.23098878860473632
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1,4,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.018889600038528444
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2,8,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.0190528005361557
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4,16,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.018681600689888
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8,32,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.0135903999209404
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16,64,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.013526399433612824
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32,128,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.013596799969673157
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,64,256,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.013593600690364837
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,128,512,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.013580800592899322
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,256,1024,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.013593600690364837
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,512,2048,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.01361600011587143
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1024,4096,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.013948799669742584
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,8,1,8,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.017103999853134155
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2048,8192,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.02678079903125763
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4096,16384,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.057948797941207886
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8192,32768,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.11934720277786255
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,16,1,16,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.00533440001308918
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16384,65536,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.2295072078704834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32768,131072,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.45414080619812014
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1,8,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.03434560000896454
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2,16,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.03424319922924042
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4,32,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.03370560109615326
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8,64,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.023654399812221526
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16,128,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.023580799996852874
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32,256,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.023632000386714935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,64,512,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.02358720004558563
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,128,1024,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.02351039946079254
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,256,2048,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.023574399948120116
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,512,4096,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.023747199773788454
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1024,8192,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.030793601274490358
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2048,16384,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.06110720038414001
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4096,32768,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.12382400035858154
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8192,65536,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.2288896083831787
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,16,1,16,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.03568960130214691
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16384,131072,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.4467520236968994
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,32,1,32,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.005654399842023849
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1,16,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.06556159853935242
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2,32,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.06508479714393615
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4,64,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.06411839723587036
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8,128,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.044038400053977966
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16,256,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.044038400053977966
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,32,512,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.04405440092086792
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,64,1024,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.04402880072593689
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,128,2048,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.0440416008234024
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,256,4096,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.043993601202964784
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,512,8192,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.048240000009536745
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1024,16384,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.06797760128974914
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2048,32768,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.1324288010597229
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4096,65536,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.2409280061721802
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8192,131072,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.453107213973999
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,32,1,32,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.06726400256156921
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,64,1,64,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.005340800061821938
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1,32,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.1264863967895508
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2,64,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.1257472038269043
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4,128,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.12495039701461792
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8,256,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.08473920226097106
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,16,512,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.08457279801368714
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,32,1024,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.08464319705963134
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,64,2048,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.08466560244560242
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,128,4096,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.0846015989780426
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,256,8192,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.08670719861984252
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,512,16384,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.09597759842872619
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1024,32768,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.14236479997634888
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2048,65536,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.2616096019744873
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4096,131072,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.48291840553283694
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,64,1,64,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.13065279722213746
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,128,1,128,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.006095999851822853
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1,64,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.24888958930969238
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2,128,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.24873600006103516
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4,256,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.24576959609985352
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,8,512,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.16605440378189087
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,16,1024,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.166048002243042
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,32,2048,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.16607999801635742
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,64,4096,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.16615359783172606
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,128,8192,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.1682047963142395
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,256,16384,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.1705664038658142
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,512,32768,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.19195519685745238
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1024,65536,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.27842559814453127
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,128,1,128,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.2597055912017822
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2048,131072,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.504636812210083
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,256,1,256,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.0084927998483181
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1,1,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.007552000135183335
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2,2,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.007356800138950348
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4,4,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.007254400104284286
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8,8,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.006032000109553337
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16,16,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0059935998171567915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32,32,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0059935998171567915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,64,64,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.005984000116586685
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,128,128,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.005977600067853928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,256,256,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.006150399893522262
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,256,1,256,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.5131999969482421
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,512,512,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.006006399914622307
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1024,1024,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.006019200012087822
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2048,2048,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.007823999971151352
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4096,4096,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.011001600325107575
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,512,1,512,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.01356160044670105
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8192,8192,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.023187200725078582
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16384,16384,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.059487998485565186
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32768,32768,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1268447995185852
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1,2,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.01117440015077591
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2,4,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.011072000116109848
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4,8,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.010982400178909302
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8,16,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.008345600217580795
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16,32,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.008390399813652038
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32,64,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.008326400071382523
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,64,128,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.00838399976491928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,128,256,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.008352000266313553
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,256,512,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.008748800307512284
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,512,1024,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.00838399976491928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1024,2048,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.008742400258779526
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2048,4096,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.01173119992017746
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4096,8192,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0245728000998497
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8192,16384,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.05630720257759094
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16384,32768,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1180799961090088
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32768,65536,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.22990720272064208
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1,4,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.01902720034122467
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2,8,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.018799999356269838
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4,16,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.018607999384403228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8,32,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.01340160071849823
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16,64,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.013510400056838989
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32,128,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.01343040019273758
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,64,256,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.013459199666976928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,128,512,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.013414399325847625
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,256,1024,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.013542400300502777
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,512,2048,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.013577599823474885
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1024,4096,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.013961599767208099
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2048,8192,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.026899200677871705
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,512,1,512,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,1.0161888122558593
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4096,16384,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.05767999887466431
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8192,32768,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.11904319524765014
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1024,1,1024,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,0.023315200209617616
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16384,65536,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.2297055959701538
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32768,131072,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.4539008140563965
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1,8,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0347680002450943
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2,16,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.03430399894714355
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4,32,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0338591992855072
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8,64,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.02369920015335083
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16,128,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.023820799589157105
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32,256,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.023798400163650514
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,64,512,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.023695999383926393
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,128,1024,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.02375040054321289
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,256,2048,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.023756800591945647
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,512,4096,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.023871999979019166
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1024,8192,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.03181119859218597
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2048,16384,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.06031680107116699
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4096,32768,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1241919994354248
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8192,65536,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.22975680828094483
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16384,131072,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.44594240188598633
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1024,1,1024,5120,4,16,128,48,128,Qwen/Qwen3.5-27B,2.029248046875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1,16,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0654304027557373
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2,32,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.06458879709243774
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1,1,1,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.005062400177121162
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1,1,1,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.005119999870657921
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4,64,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.06395519971847534
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,2,1,2,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.005455999821424484
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,2,1,2,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.005699200183153152
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,4,1,4,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0053311999887228016
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,4,1,4,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.007030399888753891
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,8,1,8,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.005286400020122528
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,8,1,8,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.010099200159311294
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8,128,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.043942400813102724
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16,256,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0438975989818573
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,16,1,16,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.005257600173354149
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,32,512,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0438400000333786
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,16,1,16,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.024911999702453613
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,64,1024,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.0438975989818573
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,32,1,32,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.004937599971890449
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,32,1,32,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.046588799357414244
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,64,1,64,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.005100800096988678
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,64,1,64,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.08824959993362427
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,128,1,128,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.006080000102519989
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,128,1,128,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.17323839664459229
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,128,2048,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.04386880099773407
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,256,4096,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.04390400052070618
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,256,1,256,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.00859839990735054
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,512,8192,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.047779199481010434
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,256,1,256,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.34336960315704346
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1024,16384,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.06756799817085266
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2048,32768,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1319615960121155
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,512,1,512,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.01345279961824417
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,512,1,512,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.680294418334961
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4096,65536,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.2422719955444336
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1024,1,1024,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.023203200101852416
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1024,1,1024,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,1.3583904266357423
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8192,131072,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.45375680923461914
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1,1,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.0072512000799179075
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1,32,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1266528010368347
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2,64,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.12643519639968873
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4,128,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.12492480278015136
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8,256,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.08467199802398681
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,16,512,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.08471999764442444
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,32,1024,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.08471999764442444
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,64,2048,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.08466240167617797
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,128,4096,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.084688001871109
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,256,8192,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.08692799806594849
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,512,16384,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.09598399996757508
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1024,32768,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1422592043876648
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2048,65536,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.26038079261779784
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4096,131072,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.4812928199768066
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2,2,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.007199999690055847
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4,4,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.007097599655389785
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8,8,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.005811199918389321
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16,16,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.0057472001761198046
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32,32,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.00597120001912117
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1,64,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.2492095947265625
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,64,64,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.00584000013768673
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,128,128,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.005859199911355972
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2,128,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.24895040988922118
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,256,256,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.005859199911355972
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,512,512,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.006028800085186958
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1024,1024,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.006121600046753883
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4,256,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.24567360877990724
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2048,2048,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.007782399654388428
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,8,512,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1660863995552063
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4096,4096,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.01127680018544197
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,16,1024,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.16597440242767333
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8192,8192,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.02359039932489395
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16384,16384,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.06382399797439575
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,32,2048,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1664512038230896
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,64,4096,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.16626559495925902
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32768,32768,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.1273151993751526
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1,2,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.011212799698114395
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,128,8192,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.16828479766845703
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,256,16384,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1705407977104187
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2,4,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.011228799819946289
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,512,32768,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.1911743998527527
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4,8,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.011084800213575363
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8,16,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.008483199775218964
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16,32,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.008524800091981888
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32,64,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.008531200140714646
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,64,128,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.008502399921417237
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1024,65536,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.2743391990661621
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,128,256,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.008579199761152267
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,256,512,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.008486399799585343
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,512,1024,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.008528000116348267
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1024,2048,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.008700799942016602
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2048,131072,2048,4,16,128,32,128,Qwen/Qwen3.5-35B-A3B,0.5003967761993409
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2048,4096,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.01175680011510849
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4096,8192,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.02449599951505661
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8192,16384,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.06305919885635376
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16384,32768,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.1037727952003479
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32768,65536,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.23069119453430176
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1,4,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.018911999464035035
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1,1,1,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.003951999917626381
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2,8,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.01886720061302185
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4,16,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.018620799481868743
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8,32,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.013468800485134125
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16,64,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.013491199910640716
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32,128,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.013439999520778656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,64,256,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.013551999628543854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,128,512,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.013478399813175201
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,256,1024,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.01348160058259964
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,512,2048,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.013523200154304504
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1024,4096,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.01387840062379837
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2048,8192,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.026643198728561402
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4096,16384,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.06508799791336059
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8192,32768,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.11945279836654663
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16384,65536,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.23025600910186766
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32768,131072,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.45447998046875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1,8,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.034643200039863584
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2,16,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.03441919982433319
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4,32,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.033897599577903746
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8,64,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.023830400407314302
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16,128,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.02377920001745224
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32,256,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.023904000222682954
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1,1,1,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.005686400085687637
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,64,512,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.02377600073814392
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,128,1024,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.023798400163650514
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,256,2048,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.023817600309848787
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,512,4096,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.024038399755954742
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1024,8192,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.03164800107479095
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,2,1,2,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.003907199949026108
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2048,16384,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.06518080234527587
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4096,32768,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.12337599992752075
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8192,65536,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.22979199886322021
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16384,131072,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.44730558395385744
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1,16,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.06575999855995178
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2,32,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.0652671992778778
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4,64,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.06437119841575623
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8,128,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.0441536009311676
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16,256,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.04419519901275635
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,32,512,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.04415999948978424
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,64,1024,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.04413439929485321
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,128,2048,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.044156798720359804
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,256,4096,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.04407039880752563
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,512,8192,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.048054400086402896
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1024,16384,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.07354559898376464
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,2,1,2,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.0068351998925209045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2048,32768,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.13193919658660888
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4096,65536,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.24285759925842285
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,4,1,4,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.003974400088191032
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8192,131072,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.45505599975585936
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1,32,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.12726720571517944
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2,64,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.12620160579681397
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4,128,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.12551039457321167
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8,256,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.08513919711112976
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,16,512,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.08511360287666321
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,32,1024,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.08526399731636047
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,64,2048,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.08518080115318298
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,128,4096,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.08508800268173218
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,256,8192,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.08708800077438354
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,512,16384,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.10096960067749024
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1024,32768,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.14287680387496948
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,4,1,4,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.00979200005531311
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2048,65536,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.26004478931427
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,8,1,8,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.004057599976658821
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4096,131072,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.4849503993988037
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1,64,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.2504832029342651
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2,128,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.2502943992614746
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4,256,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.2475264072418213
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,8,512,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.1673151969909668
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,16,1024,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.1671679973602295
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,32,2048,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.16722559928894043
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,64,4096,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.16726720333099365
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,8,1,8,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.025180798768997193
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,128,8192,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.1692512035369873
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,256,16384,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.17357120513916016
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,512,32768,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.19185919761657716
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,16,1,16,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.003977600112557411
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1024,65536,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.27514560222625734
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2048,131072,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.5143551826477051
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1,1,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.007385600358247757
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,16,1,16,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.04669440090656281
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2,2,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.007295999675989151
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4,4,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.0072672002017498015
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8,8,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.005955199897289276
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16,16,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.0060095999389886854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32,32,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.005961599946022034
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,32,1,32,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.004182400181889534
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,64,64,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.005875200033187866
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,128,128,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.005862399935722351
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,256,256,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.0058559998869895935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,512,512,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.006006399914622307
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,1024,1024,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.0059424001723527905
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,2048,2048,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.007849600166082382
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,4096,4096,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.011241599917411804
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,8192,8192,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.02375359982252121
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,16384,16384,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.06595519781112671
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,1,32768,32768,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.12744319438934326
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1,2,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.011126399785280228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2,4,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.011110399663448334
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4,8,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.010896000266075134
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8,16,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.008422400057315826
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16,32,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.008428800106048583
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32,64,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.008364800363779068
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,64,128,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.008448000252246856
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,128,256,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.008448000252246856
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,256,512,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.008476799726486206
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,512,1024,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.008499199897050858
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,1024,2048,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.008716800063848496
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,2048,4096,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.011798399686813354
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,4096,8192,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.02423679977655411
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,8192,16384,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.06350079774856568
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,16384,32768,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.1180575966835022
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,2,32768,65536,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.2304896116256714
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1,4,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.019100800156593323
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,32,1,32,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.08817920088768005
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2,8,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.019014400243759156
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4,16,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.018726399540901183
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8,32,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.01348160058259964
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16,64,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.01358720064163208
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32,128,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.013622400164604188
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,64,256,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.01345600038766861
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,64,1,64,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.004902400076389313
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,128,512,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.013494400680065155
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,256,1024,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.013494400680065155
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,512,2048,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.013577599823474885
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,1024,4096,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.013833600282669067
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,2048,8192,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.02683520019054413
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,4096,16384,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.0645792007446289
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,8192,32768,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.11871360540390015
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,16384,65536,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.2300352096557617
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,4,32768,131072,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.45391998291015623
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1,8,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.03447999954223633
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2,16,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.03427839875221252
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4,32,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.033820798993110655
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8,64,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.02356480062007904
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16,128,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.02366719990968704
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,32,256,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.02367040067911148
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,64,512,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.023820799589157105
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,128,1024,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.023839999735355378
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,256,2048,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.02378239929676056
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,512,4096,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.02391680032014847
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,1024,8192,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.03161920011043549
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,2048,16384,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.06793919801712037
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,64,1,64,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.1732192039489746
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,4096,32768,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.12337599992752075
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,128,1,128,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.006086400151252747
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,8192,65536,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.22971200942993164
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,8,16384,131072,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.44708800315856934
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1,16,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.06575679779052734
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2,32,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.06534079909324646
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4,64,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.06443840265274048
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8,128,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.044284799695014955
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,16,256,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.04423680007457733
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,32,512,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.044310399889945985
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,64,1024,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.044259199500083925
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,128,2048,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.04413439929485321
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,256,4096,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.044198399782180785
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,512,8192,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.048441600799560544
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,1024,16384,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.0748799979686737
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,2048,32768,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.13232959508895875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,4096,65536,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.24198400974273682
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,128,1,128,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.34030399322509763
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,16,8192,131072,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.4557024002075195
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,256,1,256,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.008611200004816055
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1,32,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.12711360454559326
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2,64,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.12704639434814452
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4,128,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.12550079822540283
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,8,256,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.08506879806518555
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,16,512,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.08518400192260742
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,32,1024,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.08516799807548522
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,64,2048,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.08518720269203187
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,128,4096,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.08522239923477173
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,256,8192,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.0871936023235321
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,512,16384,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.09755520224571228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,1024,32768,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.14206399917602539
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,2048,65536,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.2629055976867676
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,256,1,256,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.6763391971588135
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,32,4096,131072,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.48433599472045896
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,512,1,512,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.013551999628543854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1,64,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.2505728006362915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2,128,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.25044798851013184
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,4,256,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.24754559993743896
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,8,512,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.16737600564956664
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,16,1024,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.1673792004585266
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,32,2048,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.16738879680633545
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,64,4096,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.16736960411071777
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,128,8192,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.16922240257263182
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,256,16384,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.17350720167160033
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,512,32768,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.1921183943748474
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,1024,65536,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.2757983922958374
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,512,1,512,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,1.349343967437744
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_fn,context,64,2048,131072,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.5161183834075928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1024,1,1024,3072,4,16,128,64,128,Qwen/Qwen3.5-122B-A10B,0.023289600014686586
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1,1,1,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.0051968000829219815
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,1,1,1,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.00557439997792244
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,2,1,2,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.004944000020623207
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,2,1,2,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.0068800002336502075
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,4,1,4,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.005392000079154968
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,4,1,4,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.009683199971914292
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,8,1,8,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.005142400041222572
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,8,1,8,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.025011199712753295
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,16,1,16,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.005052800104022026
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,16,1,16,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.04623680114746094
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,32,1,32,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.005036799982190132
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,32,1,32,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.08900799751281738
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,64,1,64,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.005119999870657921
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,64,1,64,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.17435840368270875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,128,1,128,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.006054399907588959
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,128,1,128,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.3405535936355591
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,256,1,256,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.008684799820184708
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,256,1,256,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.6785823822021484
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,512,1,512,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.013606399297714233
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,fused_recurrent_gated_delta_rule,generation,512,1,512,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,1.3494239807128907
SGLang,0.5.9,NVIDIA H100 80GB HBM3,gdn,causal_conv1d_update,generation,1024,1,1024,4096,4,16,128,64,128,Qwen/Qwen3.5-397B-A17B,0.023478400707244874
