framework,version,device,op_name,kernel_source,model,architecture,mla_dtype,kv_cache_dtype,gemm_type,num_heads,batch_size,isl,tp_size,step,latency
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,0,0.9783
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,0,1.0027
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,0,1.0608
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,0,1.0315
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,0,1.1647
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,0,1.1886
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,0,1.2103
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,4,1,0,1.0179
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,0,1.1922
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,4,1,0,1.0519
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,4,1,0,1.0587
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,4,1,0,1.1280
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,4,1,0,1.0001
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,8,1,0,0.9560
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,8,1,0,1.0339
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,8,1,0,1.0210
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,4,1,0,1.1051
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,8,1,0,1.1060
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,8,1,0,1.0860
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,8,1,0,1.0374
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,4,1,0,1.1475
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,4,1,0,1.1499
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,8,1,0,1.1449
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,8,1,0,1.1256
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,16,1,0,0.9791
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,16,1,0,1.1649
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,16,1,0,1.0510
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,16,1,0,1.0404
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,16,1,0,1.1812
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,16,1,0,1.2016
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,16,1,0,1.0931
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,16,1,0,1.1708
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,32,1,0,1.0496
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,32,1,0,1.1123
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,32,1,0,1.1016
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,32,1,0,1.2174
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,32,1,0,1.2147
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,32,1,0,1.2069
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,64,1,0,1.0374
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,64,1,0,1.0598
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,64,1,0,1.0410
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,64,1,0,1.1734
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,32,1,0,1.2471
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,64,1,0,1.1849
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,64,1,0,1.1608
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,32,1,0,1.1108
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,64,1,0,1.0233
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,128,1,0,1.0316
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,128,1,0,1.0823
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,128,1,0,1.0606
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,128,1,0,1.0510
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,128,1,0,1.1889
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,128,1,0,1.1705
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,128,1,0,1.1492
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,64,1,0,1.1880
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,128,1,0,1.1859
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,256,1,0,1.0748
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,256,1,0,1.0695
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,256,1,0,1.0520
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,256,1,0,1.1733
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,256,1,0,1.2281
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,256,1,0,1.1706
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,256,1,0,1.1918
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,512,1,0,1.1111
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,256,1,0,1.0443
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,512,1,0,1.1084
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,512,1,0,1.1045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,512,1,0,1.2160
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,512,1,0,1.2475
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,512,1,0,1.2403
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,512,1,0,1.0969
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,512,1,0,1.2223
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1024,1,0,1.1315
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1024,1,0,1.1565
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1024,1,0,1.0900
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1024,1,0,1.2410
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1024,1,0,1.1847
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1024,1,0,1.3463
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1024,1,0,1.3110
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1024,1,0,1.2430
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,2048,1,0,1.0405
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,2048,1,0,1.1082
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,2048,1,0,1.2021
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,2048,1,0,1.5145
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,2048,1,0,1.1387
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,2048,1,0,1.3388
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,2048,1,0,1.6262
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,2048,1,0,1.2514
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,4096,1,0,6.4479
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,4096,1,0,6.0297
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,4096,1,0,16.1282
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,4096,1,0,16.4928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,4096,1,0,16.4935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,4096,1,0,6.4486
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,4096,1,0,16.8918
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,4096,1,0,17.3789
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,8192,1,0,6.6011
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,8192,1,0,7.1854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,8192,1,0,7.6272
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,8192,1,0,12.8183
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,8192,1,0,13.3654
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,8192,1,0,13.0768
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,8192,1,0,13.9072
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,8192,1,0,15.7023
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,16384,1,0,13.8638
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,16384,1,0,15.1429
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,16384,1,0,16.4699
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,16384,1,0,25.4021
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,16384,1,0,26.3731
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,16384,1,0,27.0513
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,16384,1,0,28.5025
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,0,0.9041
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,0,0.8845
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,4,1,0,0.8728
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,8,1,0,0.9269
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,4,1,0,0.9375
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,0,0.9189
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,16384,1,0,32.6603
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,16,1,0,0.8510
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,8,1,0,0.8726
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,4,1,0,0.8649
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,32,1,0,0.8533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,16,1,0,0.8642
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,0,0.9620
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,8,1,0,0.8719
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,64,1,0,0.8528
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,32,1,0,0.8600
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,16,1,0,0.8692
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,4,1,0,0.9675
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,128,1,0,0.8579
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,64,1,0,0.8608
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,32,1,0,0.8711
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,8,1,0,0.9613
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,256,1,0,0.8567
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,128,1,0,0.8505
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,64,1,0,0.8622
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,16,1,0,0.9641
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,512,1,0,0.8548
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,256,1,0,0.8732
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,128,1,0,0.8624
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,32,1,0,0.9519
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1024,1,0,0.8800
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,512,1,0,0.8697
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,256,1,0,0.8820
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,64,1,0,0.9624
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,2048,1,0,0.9165
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1024,1,0,0.9372
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,512,1,0,0.9286
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,128,1,0,0.9646
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,2048,1,0,1.0697
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1024,1,0,1.0267
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,256,1,0,1.0058
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,2048,1,0,1.4053
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,512,1,0,1.0812
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,0,0.9800
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1024,1,0,1.4306
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,4,1,0,1.0175
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,0,1.0070
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,0,0.8819
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,4096,1,0,6.1365
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,8,1,0,1.0206
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,4,1,0,0.8499
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,4,1,0,0.9767
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,2048,1,0,2.2091
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,16,1,0,0.9587
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,4096,1,0,6.7303
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,8,1,0,0.9354
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,0,0.9729
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,32,1,0,0.9564
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,16,1,0,0.9466
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,4,1,0,0.9580
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,64,1,0,0.9504
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,32,1,0,0.9331
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,4096,1,0,7.1490
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,8,1,0,0.9537
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,128,1,0,0.9616
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,64,1,0,0.9427
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,16,1,0,0.9550
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,256,1,0,0.9711
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,128,1,0,0.9457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,32,1,0,0.9513
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,256,1,0,0.9543
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,512,1,0,0.9577
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,64,1,0,0.9468
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,512,1,0,0.9643
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1024,1,0,0.9743
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,8,1,0,0.8468
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,128,1,0,0.9467
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1024,1,0,1.0137
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,2048,1,0,1.0179
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,16,1,0,0.8385
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,256,1,0,0.9627
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,32,1,0,0.8298
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,2048,1,0,1.1527
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,512,1,0,1.0090
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,64,1,0,0.8451
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,8192,1,0,12.5519
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,128,1,0,0.8487
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1024,1,0,1.1201
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,256,1,0,0.8869
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,2048,1,0,1.4672
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,512,1,0,0.9748
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,8192,1,0,13.7879
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,4096,1,0,15.1981
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1024,1,0,1.3076
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,8192,1,0,14.9866
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,2048,1,0,2.0879
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,4096,1,0,11.7870
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,4096,1,0,12.2792
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,4096,1,0,13.2858
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,4096,1,0,12.3260
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,8192,1,0,24.0057
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,8192,1,0,31.0617
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,8192,1,0,24.9845
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,8192,1,0,26.9015
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,8192,1,0,25.1517
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,16384,1,0,27.5842
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,16384,1,0,29.9722
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,16384,1,0,32.8713
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,16384,1,0,50.9535
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,16384,1,0,52.9314
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,16384,1,0,57.4154
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,16384,1,0,53.7278
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,0,0.9177
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,4,1,0,0.8457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,16384,1,0,67.7273
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,8,1,0,0.8433
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,16,1,0,0.8963
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,0,0.9349
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,32,1,0,0.8445
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,64,1,0,0.8477
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,4,1,0,0.9430
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,128,1,0,0.8421
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,8,1,0,0.9467
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,256,1,0,0.8499
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,0,0.8766
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,16,1,0,0.9402
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,512,1,0,0.8626
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,32,1,0,0.9361
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,0,0.8720
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1024,1,0,0.8996
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,64,1,0,0.9332
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,4,1,0,0.8489
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,128,1,0,0.9842
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,8,1,0,0.8528
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,256,1,0,1.0553
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,16,1,0,0.8538
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,32,1,0,0.8554
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,512,1,0,1.3552
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,64,1,0,1.3787
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,128,1,0,0.8788
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,4,1,0,0.8596
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1024,1,0,1.9870
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,8,1,0,0.8733
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,256,1,0,0.9377
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,16,1,0,0.8671
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,512,1,0,1.0052
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,2048,1,0,1.0803
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,32,1,0,0.8594
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1024,1,0,1.2849
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,64,1,0,0.8597
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,2048,1,0,3.8218
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,128,1,0,0.9062
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,2048,1,0,2.1754
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,256,1,0,0.8838
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,512,1,0,0.9279
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1024,1,0,1.0281
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,2048,1,0,1.4180
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,4096,1,0,11.6592
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,4096,1,0,14.0140
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,4096,1,0,12.9999
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,4096,1,0,30.1023
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,0,1.0228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,4,1,0,0.9519
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,8192,1,0,24.7754
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,8,1,0,0.9538
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,16,1,0,0.9485
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,0,1.2821
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,32,1,0,0.9573
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,64,1,0,0.9462
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,4,1,0,1.3533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,128,1,0,0.9685
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,8,1,0,1.2268
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,256,1,0,0.9535
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,16,1,0,1.2320
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,512,1,0,0.9873
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,32,1,0,1.3554
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1024,1,0,1.0122
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,64,1,0,1.3470
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,2048,1,0,1.1875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,8192,1,0,29.8011
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,128,1,0,1.3349
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,0,1.4548
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,8192,1,0,27.4855
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,0,0.8596
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,4,1,0,1.0005
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,256,1,0,1.5807
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,4,1,0,0.8702
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,8,1,0,0.9830
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,8,1,0,0.8414
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,512,1,0,1.2262
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,16,1,0,0.8389
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,16,1,0,0.9795
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,32,1,0,0.8533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1024,1,0,1.1047
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,32,1,0,0.9662
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,64,1,0,0.8770
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,64,1,0,0.9748
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,2048,1,0,1.4986
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,128,1,0,0.8980
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,128,1,0,0.9862
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,256,1,0,0.9692
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,256,1,0,1.0411
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,512,1,0,1.2695
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,512,1,0,1.1103
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1024,1,0,1.4957
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1024,1,0,1.8926
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,2048,1,0,2.2393
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,2048,1,0,3.7966
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,4096,1,0,22.8255
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,4096,1,0,24.2118
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,4096,1,0,24.2350
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,8192,1,0,64.3983
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,4096,1,0,26.0453
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,8192,1,0,47.9308
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,8192,1,0,50.1455
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,8192,1,0,52.7439
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,8192,1,0,54.3996
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,16384,1,0,58.4283
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,16384,1,0,63.4011
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,16384,1,0,69.3470
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,0,1.3585
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,4,1,0,1.3487
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,8,1,0,1.3582
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,16,1,0,1.3441
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,32,1,0,1.3453
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,64,1,0,1.3575
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,128,1,0,1.3641
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,256,1,0,1.0538
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,512,1,0,0.9109
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1024,1,0,1.0210
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,2048,1,0,1.4311
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,0,0.8745
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,0,0.8755
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,4,1,0,0.8540
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,4,1,0,0.8641
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,8,1,0,0.8718
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,8,1,0,1.0496
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,16,1,0,0.8975
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,16,1,0,0.8603
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,32,1,0,0.9636
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,32,1,0,1.0485
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,64,1,0,0.8856
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,64,1,0,0.8658
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,128,1,0,0.9028
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,128,1,0,0.8728
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,256,1,0,1.0070
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,256,1,0,0.9311
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,512,1,0,1.2304
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,512,1,0,0.9887
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1024,1,0,1.3258
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1024,1,0,1.9941
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,2048,1,0,2.2637
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,2048,1,0,3.9931
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,4096,1,0,23.1972
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,4096,1,0,25.5307
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,4096,1,0,27.7784
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,8192,1,0,57.0489
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,8192,1,0,61.3610
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,8192,1,0,64.3574
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,0,1.0000
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,4,1,0,0.9739
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,8,1,0,0.9680
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,16,1,0,0.9845
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,32,1,0,0.9984
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,64,1,0,1.0015
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,128,1,0,0.9997
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,0,0.9747
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,256,1,0,1.0000
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,4,1,0,0.9461
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,512,1,0,1.0710
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,8,1,0,0.9424
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1024,1,0,1.2682
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,16,1,0,0.9594
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,2048,1,0,1.8128
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,32,1,0,0.9813
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,64,1,0,0.9715
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,128,1,0,0.9748
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,0,0.9479
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,256,1,0,0.9716
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,4,1,0,0.9044
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,8,1,0,0.8935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,512,1,0,1.0138
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,16,1,0,0.9012
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1024,1,0,1.1206
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,32,1,0,0.9166
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,2048,1,0,1.3553
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,64,1,0,0.9247
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,128,1,0,0.9078
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,256,1,0,0.9151
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,512,1,0,0.9279
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1024,1,0,0.9740
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,2048,1,0,1.1212
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,0,1.0601
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,4,1,0,1.0292
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,8,1,0,1.0403
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,16,1,0,1.0491
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,32,1,0,1.0478
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,64,1,0,1.0598
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,128,1,0,1.0579
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,0,1.0304
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,256,1,0,1.0615
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,4,1,0,1.0109
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,512,1,0,1.1173
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,8,1,0,1.0071
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1024,1,0,1.3228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,16,1,0,1.0202
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,2048,1,0,1.8614
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,32,1,0,1.0232
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,64,1,0,1.0277
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,128,1,0,1.0293
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,256,1,0,1.0317
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,512,1,0,1.0589
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1024,1,0,1.1722
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,2048,1,0,1.4007
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,4096,1,0,4.9471
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,4096,1,0,5.1605
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,4096,1,0,4.0675
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,8192,1,0,9.8082
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,8192,1,0,8.0489
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,8192,1,0,6.8228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,16384,1,0,17.3159
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,0,0.9673
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,16384,1,0,20.6878
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,4,1,0,0.9595
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,8,1,0,0.9672
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,0,0.9860
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,16,1,0,0.9786
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,4,1,0,0.9922
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,32,1,0,0.9684
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,8,1,0,1.0004
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,64,1,0,0.9819
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,16,1,0,1.0121
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,128,1,0,0.9839
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,32,1,0,0.9950
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,256,1,0,1.0072
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,64,1,0,1.0035
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,512,1,0,1.0930
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,128,1,0,1.0046
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,256,1,0,1.0704
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1024,1,0,1.2956
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,512,1,0,1.2371
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,2048,1,0,1.8632
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1024,1,0,1.6868
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,2048,1,0,2.8847
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,4096,1,0,8.9435
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,4096,1,0,9.8943
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,16384,1,0,20.9275
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,8192,1,0,21.7747
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,0,0.9459
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,8192,1,0,26.5469
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,4,1,0,0.9559
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,8,1,0,1.1846
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,4096,1,0,8.7882
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,16,1,0,0.9426
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,32,1,0,0.9499
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,4096,1,0,7.2977
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,64,1,0,0.9407
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,128,1,0,0.9486
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,256,1,0,0.9511
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,512,1,0,0.9922
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1024,1,0,1.1044
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,2048,1,0,1.3473
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,16384,1,0,34.4162
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,4096,1,0,6.6490
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,8192,1,0,13.7860
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,0,0.9637
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,8192,1,0,16.0882
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,4,1,0,0.9662
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,8,1,0,0.9707
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,16,1,0,0.9805
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,32,1,0,0.9668
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,64,1,0,0.9735
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,128,1,0,0.9994
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,256,1,0,1.0855
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,512,1,0,1.2425
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1024,1,0,1.6985
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,8192,1,0,13.4621
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,2048,1,0,3.0564
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,16384,1,0,41.5706
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,0,0.9856
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,4,1,0,0.9879
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,8,1,0,1.0067
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,16,1,0,1.0019
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,16384,1,0,28.1470
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,32,1,0,1.0125
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,64,1,0,1.0081
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,0,1.0129
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,4096,1,0,15.5362
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,128,1,0,1.0733
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,4,1,0,1.0099
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,8,1,0,1.0130
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,256,1,0,1.2365
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,16,1,0,1.0345
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,16384,1,0,33.2522
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,512,1,0,1.6173
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,32,1,0,1.0296
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,64,1,0,1.0366
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,0,1.0427
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,128,1,0,1.0334
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,4,1,0,1.0472
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1024,1,0,2.6251
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,8,1,0,1.0340
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,256,1,0,1.0621
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,16,1,0,1.0558
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,512,1,0,1.1439
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,32,1,0,1.0836
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1024,1,0,1.3741
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,64,1,0,1.1227
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,2048,1,0,5.7742
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,128,1,0,1.0653
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,2048,1,0,1.8838
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,256,1,0,1.1254
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,512,1,0,1.2937
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1024,1,0,1.7459
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,2048,1,0,2.9137
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,4096,1,0,13.4509
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,4096,1,0,19.3597
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,8192,1,0,32.2583
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,4096,1,0,15.7949
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,8192,1,0,27.1802
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,8192,1,0,32.1452
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,8192,1,0,40.1561
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,16384,1,0,28.8571
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,16384,1,0,73.3866
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,0,0.9779
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,16384,1,0,57.0709
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,4,1,0,0.9918
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,8,1,0,0.9686
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,16,1,0,0.9507
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,32,1,0,0.9546
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,64,1,0,0.9986
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,128,1,0,1.0557
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,256,1,0,1.1903
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,512,1,0,1.6108
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1024,1,0,2.7945
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,2048,1,0,5.9598
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,16384,1,0,68.2215
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,0,0.9388
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,4,1,0,0.9351
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,8,1,0,0.9454
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,16,1,0,0.9443
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,16384,1,0,87.8007
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,4096,1,0,30.8508
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,32,1,0,0.9458
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,64,1,0,0.9413
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,0,0.9969
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,128,1,0,0.9532
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,4,1,0,1.0055
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,256,1,0,0.9976
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,8,1,0,1.0105
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,512,1,0,1.0757
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,16,1,0,1.0111
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1024,1,0,1.2616
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,32,1,0,1.0137
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,64,1,0,1.0662
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,2048,1,0,1.8594
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,128,1,0,1.1980
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,256,1,0,1.5903
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,512,1,0,2.5109
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1024,1,0,5.1654
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,0,1.0208
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,4,1,0,1.0158
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,4096,1,0,12.7934
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,8,1,0,1.0350
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,16,1,0,1.0253
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,32,1,0,1.0278
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,64,1,0,1.0341
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,128,1,0,1.0601
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,256,1,0,1.1416
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,512,1,0,1.2861
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,2048,1,0,11.3670
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1024,1,0,1.7416
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,2048,1,0,3.0371
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,0,1.0418
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,8192,1,0,26.7506
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,4,1,0,1.0420
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,8,1,0,1.0554
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,16,1,0,1.0374
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,32,1,0,1.0482
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,64,1,0,1.0631
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,128,1,0,1.5745
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,256,1,0,1.2918
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,512,1,0,1.6689
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,4096,1,0,26.6187
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1024,1,0,2.6543
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,2048,1,0,5.8673
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,4096,1,0,38.8993
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,8192,1,0,67.9863
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,4096,1,0,31.3259
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,8192,1,0,54.9266
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,8192,1,0,82.5822
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,8192,1,0,65.8014
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,16384,1,0,60.7165
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,0,0.9172
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,4,1,0,0.9319
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,8,1,0,0.9270
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,16,1,0,0.9203
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,32,1,0,0.9393
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,64,1,0,0.9426
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,128,1,0,0.9816
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,256,1,0,1.0509
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,512,1,0,1.2060
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1024,1,0,1.7043
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,2048,1,0,3.1372
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,4096,1,0,25.5586
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,8192,1,0,56.9808
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,0,1.2688
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,4,1,0,1.2391
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,8,1,0,1.2348
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,16,1,0,1.2526
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,32,1,0,1.0103
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,64,1,0,1.0115
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,128,1,0,0.9940
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,256,1,0,1.0045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,512,1,0,1.0164
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1024,1,0,1.0463
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,2048,1,0,1.2024
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,4096,1,0,6.7291
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,8192,1,0,12.6764
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,16384,1,0,25.9927
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,0,0.9967
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,4,1,0,1.0011
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,8,1,0,0.9997
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,16,1,0,1.0135
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,32,1,0,1.0765
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,64,1,0,1.0237
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,128,1,0,1.0180
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,256,1,0,1.0206
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,512,1,0,1.0465
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1024,1,0,1.1665
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,2048,1,0,1.4122
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,4096,1,0,12.4020
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,8192,1,0,31.0131
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,16384,1,0,52.6046
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,0,0.9976
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,4,1,0,0.9974
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,8,1,0,1.0342
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,16,1,0,1.0079
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,32,1,0,1.0194
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,64,1,0,1.0129
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,128,1,0,1.0160
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,256,1,0,1.2093
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,512,1,0,1.3329
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1024,1,0,1.3312
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,2048,1,0,1.9124
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,4096,1,0,33.7646
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,8192,1,0,58.1591
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,16384,1,0,136.5538
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,0,0.8855
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,4,1,0,0.8758
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,8,1,0,0.8665
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,16,1,0,0.8594
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,0,0.8338
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,32,1,0,0.8672
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,4,1,0,0.8447
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,64,1,0,0.8853
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,8,1,0,0.8482
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,0,0.8665
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,16,1,0,1.0019
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,128,1,0,1.1559
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,4,1,0,1.1155
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,32,1,0,1.1045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,8,1,0,0.8457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,64,1,0,0.8358
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,16,1,0,0.8514
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,128,1,0,0.8591
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,32,1,0,0.8921
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,256,1,0,0.9804
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,64,1,0,1.4350
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,256,1,0,2.0938
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,512,1,0,1.6494
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,128,1,0,1.5141
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,512,1,0,2.1741
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1024,1,0,2.2923
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,256,1,0,1.2329
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1024,1,0,2.0076
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,2048,1,0,4.2612
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,512,1,0,1.8172
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,2048,1,0,2.7982
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1024,1,0,3.5456
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,2048,1,0,7.9924
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,4096,1,0,61.7872
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,4096,1,0,55.8457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,4096,1,0,67.6217
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,0,0.9665
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,4,1,0,0.9746
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,8,1,0,0.9666
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,16,1,0,0.9701
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,32,1,0,0.9923
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,64,1,0,1.0738
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,128,1,0,1.2019
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,256,1,0,1.5869
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,512,1,0,2.6688
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1024,1,0,5.3707
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,2048,1,0,11.8381
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,4096,1,0,70.0039
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,0,0.9198
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,4,1,0,0.9393
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,8,1,0,0.9239
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,16,1,0,0.9317
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,32,1,0,0.9408
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,64,1,0,0.9793
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,128,1,0,1.0469
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,256,1,0,1.1928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,512,1,0,1.7544
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1024,1,0,2.8685
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,2048,1,0,6.1476
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,4096,1,0,53.7666
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,0,0.8595
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,4,1,0,0.8554
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,8,1,0,0.8504
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,16,1,0,0.8495
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,0,0.8605
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,0,0.8723
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,32,1,0,0.8728
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,4,1,0,0.8497
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,4,1,0,0.8442
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,8,1,0,0.8662
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,64,1,0,0.9132
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,8,1,0,0.8382
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,16,1,0,0.8333
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,16,1,0,0.8711
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,128,1,0,0.9641
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,32,1,0,0.8395
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,32,1,0,0.9155
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,64,1,0,0.8559
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,256,1,0,1.2302
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,64,1,0,1.3934
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,128,1,0,1.4016
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,512,1,0,2.1868
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,256,1,0,1.4861
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,128,1,0,1.6244
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,512,1,0,1.4737
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,256,1,0,1.9366
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1024,1,0,3.8388
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1024,1,0,2.3991
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,512,1,0,3.3184
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,2048,1,0,5.0823
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,2048,1,0,8.4849
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1024,1,0,7.1041
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,2048,1,0,16.3109
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,0,1.0168
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,4,1,0,0.9791
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,8,1,0,0.9801
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,16,1,0,1.0188
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,32,1,0,1.1007
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,64,1,0,1.4165
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,128,1,0,1.7343
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,256,1,0,2.5974
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,512,1,0,5.0480
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1024,1,0,10.8011
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,2048,1,0,23.6891
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,0,0.8875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,0,0.8306
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,0,0.8651
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,4,1,0,0.8898
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,4,1,0,0.8487
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,4,1,0,0.8474
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,8,1,0,0.8340
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,8,1,0,0.8448
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,16,1,0,0.8422
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,16,1,0,0.8726
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,32,1,0,0.8623
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,8,1,0,0.9169
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,32,1,0,0.9201
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,16,1,0,0.9533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,64,1,0,0.9689
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,32,1,0,1.0511
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,64,1,0,0.9002
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,128,1,0,1.2293
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,64,1,0,1.2417
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,128,1,0,1.0564
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,128,1,0,1.7930
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,256,1,0,1.9172
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,256,1,0,1.3684
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,256,1,0,3.2836
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,512,1,0,2.3056
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,512,1,0,3.6411
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1024,1,0,4.6563
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,512,1,0,6.5034
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1024,1,0,7.6311
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,2048,1,0,10.1890
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1024,1,0,14.3656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,2048,1,0,17.2314
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,2048,1,0,32.6090
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,0,1.4799
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,4,1,0,1.4478
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,8,1,0,1.2897
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,16,1,0,0.9510
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,32,1,0,0.9887
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,64,1,0,1.0437
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,128,1,0,1.1899
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,256,1,0,1.5998
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,512,1,0,2.7331
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1024,1,0,5.5734
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,2048,1,0,12.5091
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,4096,1,0,120.4939
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,0,1.0168
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,4,1,0,1.0314
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,8,1,0,1.0520
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,16,1,0,1.1315
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,32,1,0,1.2408
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,64,1,0,1.7598
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,128,1,0,2.5925
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,256,1,0,4.9991
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,512,1,0,10.1598
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1024,1,0,21.1229
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,2048,1,0,48.3479
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,0,0.8611
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,4,1,0,0.8451
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,8,1,0,0.8412
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,16,1,0,0.8549
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,32,1,0,1.1219
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,0,1.1157
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,64,1,0,1.2144
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,4,1,0,1.1029
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,128,1,0,1.4171
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,8,1,0,1.1305
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,16,1,0,1.1678
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,256,1,0,2.3001
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,32,1,0,1.2202
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,512,1,0,4.4379
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,64,1,0,1.2798
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,128,1,0,1.9329
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,0,1.0366
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,4,1,0,1.0430
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,8,1,0,1.0919
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,256,1,0,3.6093
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,16,1,0,1.1758
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,32,1,0,1.7275
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,64,1,0,2.2252
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,512,1,0,7.2192
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1024,1,0,9.0996
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,128,1,0,3.3487
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,256,1,0,6.5033
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1024,1,0,15.2811
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,512,1,0,13.3561
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1024,1,0,28.7239
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,0,1.2381
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,4,1,0,1.2331
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,8,1,0,1.5372
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,16,1,0,1.2837
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,32,1,0,1.3437
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,64,1,0,1.4791
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,128,1,0,1.8614
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,256,1,0,2.7539
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,512,1,0,5.3331
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1024,1,0,11.1850
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,2048,1,0,24.8070
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,0,1.1528
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,4,1,0,1.1222
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,8,1,0,1.1916
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,16,1,0,1.3199
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,32,1,0,1.6596
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,64,1,0,2.6013
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,128,1,0,4.8829
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,256,1,0,9.8161
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,512,1,0,20.0487
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1024,1,0,43.3171
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,0,1.6199
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,4,1,0,1.2599
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,0,0.9922
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,8,1,0,1.4927
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,4,1,0,1.0366
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,8,1,0,1.1003
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,16,1,0,1.5091
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,16,1,0,1.1626
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,32,1,0,1.0415
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,32,1,0,1.3741
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,64,1,0,1.2772
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,64,1,0,1.9953
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,128,1,0,2.2928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,128,1,0,3.5828
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,256,1,0,4.3993
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,256,1,0,7.0375
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,512,1,0,8.7680
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,0,1.2263
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,4,1,0,1.7506
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,8,1,0,1.9304
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,512,1,0,14.4825
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,16,1,0,1.5795
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,32,1,0,2.0682
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,64,1,0,3.4005
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,128,1,0,6.3915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,256,1,0,13.0699
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,512,1,0,26.6011
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,0,1.1996
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,4,1,0,1.0024
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,8,1,0,1.0389
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,16,1,0,1.1053
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,32,1,0,1.2370
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,64,1,0,1.6001
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,128,1,0,2.6670
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,256,1,0,5.1979
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,512,1,0,10.6875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1024,1,0,22.4322
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,0,1.3003
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,4,1,0,1.3543
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,8,1,0,1.4821
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,16,1,0,1.8282
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,32,1,0,2.7466
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,64,1,0,4.9632
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,128,1,0,9.8207
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,256,1,0,20.1723
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,512,1,0,40.4761
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,0,0.9935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,0,1.2393
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,4,1,0,1.0667
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,8,1,0,1.0940
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,4,1,0,1.3255
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,16,1,0,1.1824
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,8,1,0,1.3984
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,32,1,0,1.3974
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,16,1,0,1.6094
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,64,1,0,2.3546
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,32,1,0,2.1903
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,128,1,0,4.4073
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,64,1,0,3.7082
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,256,1,0,8.6775
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,128,1,0,7.1583
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,0,1.9791
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,256,1,0,14.2708
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,4,1,0,1.8779
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,8,1,0,2.1067
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,16,1,0,2.9350
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,32,1,0,3.8520
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,64,1,0,6.7919
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,128,1,0,13.0043
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,256,1,0,26.2243
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,0,1.0706
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,4,1,0,1.1198
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,8,1,0,1.1709
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,16,1,0,1.3051
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,32,1,0,1.6719
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,64,1,0,2.6839
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,128,1,0,5.1696
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,256,1,0,10.3989
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,512,1,0,21.3314
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,0,1.5916
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,4,1,0,1.8110
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,8,1,0,2.1406
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,16,1,0,3.0695
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,32,1,0,5.2757
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,64,1,0,9.8076
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,128,1,0,19.7706
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,256,1,0,39.7860
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,0,1.2568
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,4,1,0,1.3580
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,8,1,0,1.4301
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,16,1,0,1.8240
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,0,1.7236
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,32,1,0,2.5834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,4,1,0,1.8768
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,8,1,0,2.0993
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,16,1,0,2.6541
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,64,1,0,4.5532
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,128,1,0,8.6384
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,32,1,0,4.2000
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,0,2.6479
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,64,1,0,7.3215
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,4,1,0,3.0389
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,8,1,0,3.6412
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,128,1,0,14.0298
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,16,1,0,4.8213
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,32,1,0,7.5615
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,64,1,0,13.4606
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,128,1,0,26.0850
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,0,1.2433
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,4,1,0,1.3416
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,8,1,0,1.4682
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,16,1,0,1.8258
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,32,1,0,2.8324
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,64,1,0,5.2391
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,128,1,0,10.2115
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,256,1,0,20.6619
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,0,2.3374
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,4,1,0,2.8017
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,8,1,0,3.6963
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,16,1,0,5.9561
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,32,1,0,10.4681
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,64,1,0,19.6785
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,128,1,0,39.4434
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,0,1.5800
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,4,1,0,1.8383
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,8,1,0,2.1819
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,16,1,0,3.1589
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,32,1,0,5.5227
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,64,1,0,10.3359
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_context_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,128,1,0,20.6824
