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_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,1,0.0865
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,4,0.0862
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,8,0.0863
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,16,0.0859
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,32,0.0861
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,64,0.0868
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,128,0.0872
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,256,0.0887
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,512,0.0911
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,1024,0.0950
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,2048,0.1059
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,4096,0.1134
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,8192,0.1179
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,1,0.1470
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,16384,0.1278
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,4,0.1474
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,8,0.1475
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,16,0.1477
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,32,0.1480
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,64,0.1508
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,1,0.1134
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,4,0.1134
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,8,0.1130
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,16,0.1129
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,32,0.1136
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,64,0.1163
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,128,0.1162
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,256,0.1169
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,512,0.1178
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,1024,0.1192
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,1,0.0872
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,2048,0.1296
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,4,0.0872
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,4096,0.1333
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,8,0.0877
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,8192,0.1382
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,16,0.0872
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,128,0.1517
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,32,0.0872
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,16384,0.1466
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,64,0.0874
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,128,0.0882
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,256,0.0888
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,512,0.0910
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,1024,0.0935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,2048,0.1048
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,256,0.1515
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,4096,0.1088
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,8192,0.1141
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,16384,0.1240
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,512,0.1525
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,1,0.1152
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,1,0.0905
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,4,0.1150
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,4,0.0906
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,8,0.1147
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,8,0.0904
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,16,0.1147
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,16,0.0906
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,32,0.1143
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,32,0.0908
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,64,0.1181
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,1024,0.1548
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,64,0.0908
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,128,0.1185
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,128,0.0911
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,256,0.1189
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,256,0.0924
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,512,0.1195
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,512,0.0937
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,1024,0.1220
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,1024,0.0959
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,2048,0.1303
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,2048,0.1072
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,4096,0.1341
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,4096,0.1114
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,8192,0.1383
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,8192,0.1173
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,2048,0.1591
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,16384,0.1481
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,16384,0.1278
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,1,0.1143
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,4,0.1146
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,8,0.1143
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,16,0.1144
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,32,0.1146
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,64,0.1177
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,1,0.0945
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,128,0.1176
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,4,0.0949
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,256,0.1184
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,8,0.0950
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,4096,0.1680
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,512,0.1206
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,16,0.0949
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,8192,0.1724
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,1024,0.1249
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,32,0.0945
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,2048,0.1312
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,64,0.0947
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1,1,1,16384,0.1820
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,128,0.0962
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,4096,0.1374
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,256,0.0980
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,8192,0.1428
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,512,0.1003
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,1024,0.1033
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,16384,0.1536
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,2048,0.1123
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,1,0.1154
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,4096,0.1167
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,4,0.1151
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,8,0.1166
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,8192,0.1222
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,16,0.1153
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,32,0.1156
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,16384,0.1354
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,64,0.1197
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,128,0.1213
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,256,0.1220
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,512,0.1261
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,1024,0.1285
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,1,0.0970
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,2048,0.1407
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,1,0.1029
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,4,0.0980
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,4,0.1029
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,4096,0.1450
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,8,0.0979
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,8,0.1036
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,16,0.0978
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,8192,0.1504
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,16,0.1037
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,32,0.0977
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,32,0.1038
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,64,0.0997
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,64,0.1048
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,16384,0.1625
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,128,0.1000
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,128,0.1053
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,256,0.1009
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,1,0.1167
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,512,0.1006
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,4,0.1168
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,1024,0.1015
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,8,0.1173
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,2048,0.1109
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,4096,0.1145
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,8192,0.1188
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,1,0.1179
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,16384,0.1282
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,4,0.1185
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,8,0.1177
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,16,0.1183
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,256,0.1077
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,32,0.1179
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,512,0.1113
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,64,0.1184
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,1024,0.1129
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,128,0.1190
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,1,0.0991
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,1,0.0818
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,1,0.1537
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,256,0.1203
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,16,0.1173
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,2048,0.1230
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,4,0.0998
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,4,0.0815
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,4,0.1534
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,512,0.1219
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,8,0.0818
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,8,0.0999
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,32,0.1174
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,4096,0.1276
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,8,0.1531
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,1024,0.1264
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,16,0.0817
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,16,0.0991
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,64,0.1233
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,16,0.1539
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,2048,0.1411
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,32,0.0998
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,128,0.1257
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,8192,0.1356
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,32,0.1533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,4096,0.1447
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,64,0.1000
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,256,0.1275
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,1,0.1478
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,64,0.1539
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,8192,0.1500
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,512,0.1306
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,4,0.1476
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,1,0.0976
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,128,0.1544
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,16384,0.1592
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,8,0.1475
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,1024,0.1346
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,4,0.0984
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,256,0.1558
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,16,0.1477
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,8,0.0977
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,2048,0.1512
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,512,0.1573
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,32,0.1474
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,1024,0.1619
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,64,0.1510
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,4096,0.1559
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,2048,0.1771
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,128,0.1509
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,256,0.1524
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,4096,0.1820
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,8192,0.1635
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,32,0.0820
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,512,0.1538
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,8192,0.1869
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,64,0.0827
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,1024,0.1575
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,128,0.0832
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,128,0.1004
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1,1,1,16384,0.1964
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,2048,0.1619
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,256,0.0834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,256,0.1022
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,4096,0.1688
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,512,0.0852
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,512,0.1037
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,8192,0.1732
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,1024,0.0862
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,1024,0.1078
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,16,0.0981
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,2048,0.0924
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,2,1,1,16384,0.1838
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,2048,0.1187
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,32,0.0983
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,4096,0.1005
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,1,0.1208
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,4096,0.1258
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,64,0.1005
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,8192,0.1051
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,4,0.1206
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,8192,0.1306
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,128,0.1004
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,16384,0.1146
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,8,0.1208
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,256,0.1007
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,16384,0.1401
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,16,0.1208
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,512,0.1014
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,32,0.1208
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,1024,0.1029
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,1,0.1493
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,64,0.1208
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,2048,0.1086
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,4,0.1504
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,128,0.1220
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,4096,0.1150
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,8,0.1504
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,256,0.1228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,8192,0.1205
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,16,0.1504
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,512,0.1241
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,16384,0.1303
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,32,0.1499
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,1,0.1562
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,1024,0.1262
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,64,0.1547
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,2048,0.1347
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,4,0.1558
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,128,0.1553
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,4096,0.1411
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,8,0.1564
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,256,0.1568
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,8192,0.1460
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,16,0.1562
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,512,0.1601
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,32,0.1560
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,16384,0.1559
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,1024,0.1620
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,64,0.1559
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,1,0.1219
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,2048,0.1724
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,128,0.1570
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,1,0.0821
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,4,0.1223
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,1,0.1008
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,4096,0.1757
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,4,0.0825
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,256,0.1574
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,8,0.1223
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,4,0.1007
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,8,0.0823
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,512,0.1587
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,8192,0.1806
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,16,0.1222
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,8,0.1012
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,16,0.0830
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,1024,0.1611
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,32,0.1227
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,16,0.1010
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,32,0.0829
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,1,0.0980
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,4,1,1,16384,0.1918
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,2048,0.1725
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,64,0.1229
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,32,0.1010
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,64,0.0848
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,4,0.0982
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,128,0.1236
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,64,0.1009
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,8,0.0987
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,256,0.1243
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,128,0.1019
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,16,0.0983
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,256,0.1028
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,32,0.0982
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,512,0.1044
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,64,0.1006
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,1024,0.1061
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,128,0.1005
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,1,0.1511
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,2048,0.1176
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,256,0.1008
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,4,0.1502
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,512,0.1022
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,8,0.1514
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,1024,0.1040
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,16,0.1511
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,2048,0.1145
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,32,0.1509
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,4096,0.1180
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,4096,0.1754
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,64,0.1555
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,8192,0.1225
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,128,0.1566
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,8192,0.1812
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,512,0.1252
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,16384,0.1336
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,256,0.1602
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,2,1,1,16384,0.1909
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,1024,0.1265
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,512,0.1628
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,1,0.0977
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,2048,0.1375
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,1,0.1608
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,4,0.0983
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,1024,0.1656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,4096,0.1410
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,8,0.0977
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,4,0.1612
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,4096,0.1207
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,2048,0.1782
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,16,0.0983
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,8192,0.1462
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,8,0.1614
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,8192,0.1257
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,4096,0.1830
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,32,0.0987
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,16,0.1616
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,64,0.1010
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,16384,0.1565
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,16384,0.1357
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,32,0.1613
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,8192,0.1889
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,128,0.1008
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,128,0.0850
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,1,0.1288
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,64,0.1609
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,1,0.1034
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,256,0.1022
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,256,0.0854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,4,0.1294
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,4,0.1032
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,512,0.1045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,512,0.0864
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,128,0.1611
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,8,1,1,16384,0.2007
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,8,0.1291
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,8,0.1032
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,1024,0.0883
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,256,0.1621
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,16,0.1289
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,16,0.1034
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,2048,0.0953
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,512,0.1631
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,32,0.1033
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,4096,0.1024
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,1024,0.1643
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,64,0.1037
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,8192,0.1082
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,128,0.1043
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,2048,0.1745
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,16384,0.1169
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,256,0.1057
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,4096,0.1782
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,1,0.1531
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,512,0.1062
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,1,0.0834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,4,0.1532
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,8192,0.1829
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,1024,0.1082
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,4,0.0837
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,8,0.1533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,8,0.0837
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,4,1,1,16384,0.1940
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,16,0.0834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,16,0.1533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,32,0.0835
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,1024,0.1074
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,32,0.1546
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,1,0.1692
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,64,0.0852
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,2048,0.1188
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,64,0.1618
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,32,0.1292
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,128,0.0854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,4096,0.1234
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,128,0.1651
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,256,0.0861
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,64,0.1297
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,512,0.0881
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,256,0.1667
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,128,0.1301
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,8192,0.1290
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,1024,0.0935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,256,0.1313
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,512,0.1701
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,2048,0.1043
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,512,0.1325
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,1024,0.1755
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,16384,0.1415
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,4096,0.1081
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,1024,0.1339
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,2048,0.1184
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,2048,0.1932
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,8192,0.1129
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,1,0.0988
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,2048,0.1427
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,4096,0.1220
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,4096,0.1972
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,16384,0.1242
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,4096,0.1468
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,8192,0.1272
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,1,0.0830
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,4,0.1692
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,8192,0.1528
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,8192,0.2046
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,16384,0.1382
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,4,0.0829
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,8,0.1694
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,8,0.0820
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,1,0.1092
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,16,0.1693
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,16,0.0828
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,16384,0.1651
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,4,0.1091
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,32,0.0829
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,32,0.1688
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,8,0.1090
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,64,0.0854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,1,0.1394
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,64,0.1694
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,16,0.1095
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,128,0.0867
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,4,0.1396
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,32,0.1092
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,128,0.1699
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,256,0.0892
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,8,0.1396
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,64,0.1093
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,256,0.1692
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,512,0.0932
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,4,0.0993
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,128,0.1099
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,16,0.1400
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,1024,0.0967
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,512,0.1705
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,8,0.0993
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,256,0.1108
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,32,0.1397
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,2048,0.1075
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,1024,0.1723
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,16,0.0995
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,512,0.1124
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,64,0.1404
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,4096,0.1124
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,32,0.0993
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,2048,0.1821
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,1024,0.1137
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,128,0.1408
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,64,0.1025
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,8192,0.1178
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,2048,0.1229
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,4096,0.1866
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,256,0.1416
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,128,0.1035
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,4096,0.1270
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,512,0.1434
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,256,0.1051
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,8192,0.1934
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,16384,0.1304
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,8192,0.1331
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,512,0.1081
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,1024,0.1439
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,1024,0.1122
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,1,0.0858
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,2048,0.1545
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,8,1,1,16384,0.2050
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,4,0.0854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,16384,0.1451
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,2048,0.1277
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,4096,0.1595
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,8,0.0851
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,1,0.1803
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,4096,0.1332
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,1,0.1182
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,16,0.0854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,4,0.1809
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,8192,0.1676
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,32,0.0864
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,8192,0.1399
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,8,0.1806
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,64,0.0901
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,128,0.0919
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,16,0.1803
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,256,0.0940
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,32,0.1815
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,512,0.0967
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,64,0.1809
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,1024,0.1003
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,128,0.1814
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,2048,0.1158
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,256,0.1809
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,4096,0.1207
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,4,0.1182
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,8192,0.1288
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,8,0.1181
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,16,0.1184
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,32,0.1182
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,64,0.1185
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,128,0.1194
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,256,0.1199
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,512,0.1220
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,512,0.1822
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,1024,0.1234
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,1024,0.1840
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,2048,0.1336
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,2048,0.1967
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,4096,0.1384
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,4096,0.2009
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,8192,0.1462
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,8192,0.2091
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,16384,0.1510
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,16,1,1,16384,0.2255
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,16,1,1,16384,0.2197
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,16384,0.1831
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,16384,0.1807
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,16384,0.1625
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,16384,0.1563
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,16384,0.1449
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,1,0.1255
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,4,0.1261
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,8,0.1256
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,16,0.1256
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,32,0.1260
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,64,0.1273
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,128,0.1284
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,256,0.1298
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,512,0.1325
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,1024,0.1375
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,2048,0.1482
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,4096,0.1548
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,1,0.1601
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,4,0.1597
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,8,0.1600
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,16,0.1602
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,32,0.1602
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,64,0.1626
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,128,0.1754
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,256,0.1785
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,512,0.1829
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,1024,0.1918
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,2048,0.2164
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,32,1,1,4096,0.2230
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,1,0.1029
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,4,0.1027
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,8,0.1026
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,16,0.1030
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,32,0.1034
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,64,0.1069
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,128,0.1095
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,256,0.1125
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,512,0.1154
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,1024,0.1242
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,2048,0.1459
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,4096,0.1535
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,1,0.1230
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,4,0.1226
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,8,0.1229
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,16,0.1227
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,32,0.1231
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,64,0.1312
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,128,0.1348
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,256,0.1368
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,512,0.1405
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,1024,0.1505
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,2048,0.1738
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,4096,0.1806
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,1,0.1616
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,4,0.1617
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,8,0.1619
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,16,0.1617
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,32,0.1619
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,64,0.1624
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,128,0.1630
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,256,0.1644
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,512,0.1665
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,1024,0.1705
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,2048,0.1810
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,4096,0.1884
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,1,0.0910
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,4,0.0906
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,8,0.0907
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,16,0.0907
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,32,0.0914
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,64,0.0963
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,128,0.0976
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,256,0.1004
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,512,0.1042
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,1024,0.1121
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,2048,0.1341
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,4096,0.1411
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,1,0.2018
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,4,0.2021
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,8,0.2015
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,16,0.2019
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,32,0.2014
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,64,0.2011
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,128,0.2022
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,256,0.2041
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,512,0.2052
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,1024,0.2098
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,2048,0.2217
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,32,1,1,4096,0.2286
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,1,0.1395
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,4,0.1395
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,8,0.1398
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,16,0.1404
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,32,0.1401
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,64,0.1408
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,128,0.1409
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,256,0.1426
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,512,0.1439
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,1024,0.1487
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,2048,0.1592
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,4096,0.1656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,1,0.1547
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,4,0.1542
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,8,0.1547
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,16,0.1543
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,32,0.1550
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,64,0.1555
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,128,0.1566
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,256,0.1578
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,512,0.1612
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,1024,0.1661
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,2048,0.1853
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,4096,0.1952
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,1,0.1704
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,4,0.1692
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,8,0.1698
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,16,0.1696
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,32,0.1703
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,64,0.1733
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,128,0.1751
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,256,0.1785
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,512,0.1869
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,1024,0.2033
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,2048,0.2418
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,64,1,1,4096,0.2527
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,1,0.1293
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,4,0.1287
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,8,0.1291
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,16,0.1292
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,32,0.1293
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,64,0.1320
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,128,0.1443
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,256,0.1489
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,512,0.1578
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,1024,0.1730
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,2048,0.2091
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,4096,0.2192
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,1,0.1074
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,4,0.1073
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,8,0.1076
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,16,0.1079
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,32,0.1080
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,64,0.1110
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,128,0.1187
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,256,0.1211
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,512,0.1301
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,1024,0.1453
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,2048,0.1811
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,4096,0.1902
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,1,0.0957
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,4,0.0953
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,8,0.0957
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,16,0.0960
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,32,0.0966
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,64,0.1000
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,128,0.1041
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,256,0.1085
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,512,0.1166
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,1024,0.1313
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,2048,0.1666
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,4096,0.1770
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,1,0.1887
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,4,0.1886
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,8,0.1890
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,16,0.1891
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,32,0.1893
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,64,0.1900
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,128,0.1906
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,256,0.1915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,512,0.1937
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,1024,0.1978
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,2048,0.2164
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,4096,0.2263
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,1,0.2299
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,4,0.2301
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,8,0.2299
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,16,0.2302
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,32,0.2293
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,64,0.2300
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,128,0.2307
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,256,0.2316
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,512,0.2332
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,1024,0.2377
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,2048,0.2570
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,64,1,1,4096,0.2675
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,1,0.1662
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,4,0.1661
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,8,0.1662
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,16,0.1662
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,32,0.1666
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,64,0.1669
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,128,0.1677
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,256,0.1689
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,512,0.1714
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,1024,0.1762
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,2048,0.1959
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,4096,0.2054
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,1,0.2289
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,4,0.2295
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,8,0.2297
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,16,0.2297
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,32,0.2297
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,64,0.2312
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,128,0.2327
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,256,0.2358
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,512,0.2416
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,1024,0.2527
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,2048,0.2839
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,1,0.1974
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,4,0.1986
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,8,0.1981
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,16,0.1983
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,32,0.2002
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,64,0.2043
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,128,0.2094
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,256,0.2190
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,512,0.2375
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,1024,0.2703
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,128,1,1,2048,0.3395
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,1,0.1190
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,4,0.1195
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,8,0.1191
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,16,0.1195
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,32,0.1205
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,64,0.1244
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,128,0.1278
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,256,0.1362
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,512,0.1509
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,1024,0.1792
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,2048,0.2427
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,1,0.1448
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,4,0.1450
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,8,0.1450
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,16,0.1452
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,32,0.1451
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,64,0.1494
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,128,0.1531
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,256,0.1608
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,512,0.1765
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,1024,0.2042
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,2048,0.2688
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,1,0.1063
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,4,0.1064
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,8,0.1065
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,16,0.1066
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,32,0.1073
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,64,0.1110
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,128,0.1154
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,256,0.1232
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,512,0.1386
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,1024,0.1663
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,2048,0.2301
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,1,0.2705
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,4,0.2707
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,8,0.2704
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,16,0.2710
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,32,0.2711
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,64,0.2720
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,128,0.2729
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,256,0.2740
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,512,0.2794
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,1024,0.2909
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,2048,0.3222
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,1,0.3181
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,4,0.3180
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,8,0.3184
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,16,0.3184
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,32,0.3177
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,64,0.3195
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,128,0.3207
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,256,0.3241
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,512,0.3298
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,1024,0.3417
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,128,1,1,2048,0.3719
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,1,0.2433
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,4,0.2436
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,8,0.2435
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,16,0.2439
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,32,0.2439
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,64,0.2456
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,128,0.2474
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,256,0.2500
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,512,0.2550
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,1024,0.2663
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,2048,0.2971
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,1,0.3796
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,4,0.3799
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,8,0.3803
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,16,0.3804
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,32,0.3810
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,64,0.3837
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,128,0.3868
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,256,0.3920
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,512,0.4037
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,1024,0.4267
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,1,0.2665
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,4,0.2672
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,8,0.2692
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,16,0.2692
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,32,0.2707
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,64,0.2771
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,128,0.2861
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,256,0.3037
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,512,0.3390
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,256,1,1,1024,0.4034
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,1,0.1475
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,4,0.1478
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,8,0.1476
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,16,0.1481
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,32,0.1495
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,64,0.1551
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,128,0.1640
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,256,0.1815
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,512,0.2120
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,1024,0.2677
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,1,0.1796
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,4,0.1802
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,8,0.1800
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,16,0.1819
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,32,0.1835
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,64,0.1905
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,128,0.1985
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,256,0.2160
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,512,0.2468
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,1024,0.3037
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,1,0.1307
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,4,0.1310
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,8,0.1314
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,16,0.1318
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,32,0.1326
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,64,0.1392
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,128,0.1475
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,256,0.1650
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,512,0.1943
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,1024,0.2507
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,1,0.4304
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,4,0.4306
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,8,0.4314
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,16,0.4318
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,32,0.4333
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,64,0.4341
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,128,0.4365
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,256,0.4426
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,512,0.4552
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,1024,0.4786
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,1,0.5128
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,4,0.5130
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,8,0.5112
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,16,0.5121
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,32,0.5135
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,64,0.5159
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,128,0.5191
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,256,0.5257
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,512,0.5380
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,256,1,1,1024,0.5625
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,1,0.3971
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,4,0.3974
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,8,0.3979
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,16,0.3978
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,32,0.3993
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,64,0.4014
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,128,0.4041
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,256,0.4073
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,512,0.4201
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,1024,0.4432
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,1,0.6845
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,4,0.6859
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,8,0.6862
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,16,0.6809
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,32,0.6820
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,64,0.6853
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,128,0.6919
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,256,0.7049
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,512,0.7291
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,512,1,1,1,0.4412
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,512,1,1,4,0.4434
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,512,1,1,8,0.4447
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,512,1,1,16,0.4457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,512,1,1,32,0.4482
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,512,1,1,64,0.4576
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,512,1,1,128,0.4713
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,512,1,1,256,0.5062
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,512,1,1,512,0.5758
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,1,0.2173
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,4,0.2187
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,8,0.2190
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,16,0.2213
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,32,0.2251
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,64,0.2364
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,128,0.2514
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,256,0.2853
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,512,0.3441
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,1,0.2866
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,4,0.2879
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,8,0.2879
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,16,0.2905
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,32,0.2932
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,64,0.3017
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,128,0.3175
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,256,0.3521
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,512,0.4118
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,1,0.1896
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,4,0.1905
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,8,0.1904
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,16,0.1912
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,32,0.1935
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,64,0.2068
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,128,0.2212
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,256,0.2546
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,512,0.3135
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,1,0.7878
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,4,0.7883
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,8,0.7884
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,16,0.7846
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,32,0.7858
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,64,0.7885
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,128,0.7955
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,256,0.8090
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,512,0.8346
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,512,1,1,1,0.9233
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,512,1,1,4,0.9235
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,512,1,1,8,0.9170
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,512,1,1,16,0.9180
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,512,1,1,32,0.9193
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,512,1,1,64,0.9213
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,512,1,1,128,0.9280
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,512,1,1,256,0.9414
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,512,1,1,512,0.9672
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,1,0.7134
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,4,0.7141
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,8,0.7151
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,16,0.7116
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,32,0.7145
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,64,0.7166
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,128,0.7228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,256,0.7350
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,512,0.7603
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,1,1.2804
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,4,1.2763
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,8,1.2729
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,16,1.2733
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,32,1.2872
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,64,1.2841
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,128,1.2972
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,256,1.3232
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1024,1,1,1,0.8048
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1024,1,1,4,0.8116
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1024,1,1,8,0.8177
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1024,1,1,16,0.8181
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1024,1,1,32,0.8207
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1024,1,1,64,0.8458
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1024,1,1,128,0.8708
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,128,1024,1,1,256,0.9296
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,1,0.3607
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,4,0.3615
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,8,0.3618
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,16,0.3643
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,32,0.3673
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,64,0.3854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,128,0.4123
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,256,0.4735
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,1,0.4906
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,4,0.4958
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,8,0.4960
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,16,0.4986
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,32,0.5003
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,64,0.5152
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,128,0.5408
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,256,0.6059
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,1,0.2291
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,4,0.2289
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,8,0.2292
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,16,0.2293
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,32,0.2286
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,64,0.2320
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,128,0.2325
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,256,0.2326
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,512,0.2326
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,1024,0.2337
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,2048,0.2392
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,4096,0.2482
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,8192,0.2512
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1,1,1,16384,0.2606
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,1,0.2319
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,4,0.2323
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,8,0.2323
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,16,0.2325
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,32,0.2318
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,64,0.2355
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,128,0.2355
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,256,0.2358
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,512,0.2361
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,1024,0.2394
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,2048,0.2442
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,4096,0.2522
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,8192,0.2570
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,2,1,1,16384,0.2667
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,1,0.2360
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,4,0.2351
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,8,0.2355
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,16,0.2354
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,32,0.2358
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,64,0.2389
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,128,0.2392
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,256,0.2397
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,512,0.2423
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,1024,0.2472
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,2048,0.2571
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,4096,0.2605
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,8192,0.2665
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,4,1,1,16384,0.2772
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,1,0.2457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,4,0.2456
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,8,0.2457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,16,0.2452
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,32,0.2457
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,64,0.2503
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,128,0.2506
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,256,0.2526
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,512,0.2556
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,1024,0.2585
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,2048,0.2702
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,4096,0.2750
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,8192,0.2811
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,8,1,1,16384,0.2929
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,1,0.2614
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,4,0.2619
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,8,0.2616
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,16,0.2611
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,32,0.2613
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,64,0.2676
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,128,0.2695
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,256,0.2732
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,512,0.2756
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,1024,0.2801
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,2048,0.2971
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,4096,0.3024
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,8192,0.3098
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,16,1,1,16384,0.3254
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,1,0.2607
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,4,0.2599
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,8,0.2605
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,16,0.2598
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,32,0.2603
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,64,0.2686
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,128,0.2726
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,256,0.2748
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,512,0.2782
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,1024,0.2885
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,2048,0.3119
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,32,1,1,4096,0.3188
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,1,0.2701
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,4,0.2708
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,8,0.2707
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,16,0.2713
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,32,0.2714
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,64,0.2740
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,128,0.2863
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,256,0.2913
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,512,0.3000
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,1024,0.3149
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,2048,0.3518
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,64,1,1,4096,0.3609
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,1,0.2925
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,4,0.2923
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,8,0.2922
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,16,0.2928
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,32,0.2932
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,64,0.2967
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,128,0.2996
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,256,0.3075
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,512,0.3232
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,1024,0.3512
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,128,1,1,2048,0.4156
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,1,0.3309
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,4,0.3314
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,8,0.3315
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,16,0.3329
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,32,0.3345
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,64,0.3406
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,128,0.3486
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,256,0.3666
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,512,0.3976
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,256,1,1,1024,0.4537
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,1,0.4885
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,4,0.4885
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,8,0.4875
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,16,0.4895
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,32,0.4921
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,64,0.5020
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,128,0.5166
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,256,0.5511
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,512,1,1,512,0.6113
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,1,0.8280
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,4,0.8343
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,8,0.8338
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,16,0.8355
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,1,0.1757
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,4,0.1755
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,8,0.1760
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,16,0.1770
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,32,0.8354
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,32,0.1759
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,64,0.1790
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,128,0.1790
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,256,0.1794
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,64,0.8540
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,512,0.1794
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,1024,0.1799
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,2048,0.1847
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,4096,0.1924
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,128,0.8787
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,8192,0.1957
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1,1,1,16384,0.2048
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,1,0.1777
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,4,0.1777
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,8,0.1787
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,16,0.1779
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,32,0.1778
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,64,0.1806
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,128,0.1799
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,256,0.1804
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,512,0.1810
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,1024,0.1826
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,2048,0.1915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,4096,0.1959
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,8192,0.2005
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,2,1,1,16384,0.2104
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,1,0.1812
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,4,0.1810
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,8,0.1810
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,16,0.1814
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,32,0.1812
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,64,1024,1,1,256,0.9412
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,64,0.1833
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,128,0.1834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,256,0.1833
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,512,0.1851
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,1024,0.1868
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,2048,0.1943
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,4096,0.2013
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,8192,0.2066
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,4,1,1,16384,0.2174
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,1,0.1895
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,4,0.1895
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,8,0.1899
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,16,0.1895
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,32,0.1897
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,64,0.1915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,128,0.1930
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,256,0.1938
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,512,0.1956
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,1024,0.1985
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,2048,0.2098
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,4096,0.2142
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,8192,0.2205
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,8,1,1,16384,0.2328
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,1,0.2054
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,4,0.2058
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,8,0.2052
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,16,0.2051
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,32,0.2053
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,64,0.2086
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,128,0.2108
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,256,0.2120
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,512,0.2145
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,1024,0.2192
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,2048,0.2348
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,4096,0.2395
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,8192,0.2477
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,16,1,1,16384,0.2633
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,1,0.2011
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,4,0.2010
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,8,0.2022
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,16,0.2020
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,32,0.2010
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,64,0.2067
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,128,0.2086
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,256,0.2117
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,512,0.2152
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,1024,0.2238
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,2048,0.2459
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,32,1,1,4096,0.2531
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,1,0.2079
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,4,0.2082
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,8,0.2078
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,16,0.2073
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,32,0.2081
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,64,0.2113
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,128,0.2195
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,256,0.2224
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,512,0.2312
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,1024,0.2468
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,2048,0.2811
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,64,1,1,4096,0.2925
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,1,0.2203
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,4,0.2212
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,1,0.2946
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,8,0.2211
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,16,0.2210
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,32,0.2213
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,4,0.2963
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,64,0.2253
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,128,0.2285
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,256,0.2363
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,8,0.2974
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,512,0.2515
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,16,0.3006
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,1024,0.2799
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,128,1,1,2048,0.3432
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,32,0.3040
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,1,0.2540
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,4,0.2534
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,8,0.2537
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,16,0.2544
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,32,0.2554
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,64,0.2614
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,64,0.3248
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,128,0.2686
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,256,0.2855
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,128,0.3514
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,512,0.3173
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,256,1,1,1024,0.3719
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,256,0.4124
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,1,0.3433
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,4,0.3438
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,8,0.3444
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,16,0.3463
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,32,0.3488
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,64,0.3593
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,128,0.3751
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,256,0.4088
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,1,0.1495
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,4,0.1490
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,8,0.1498
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,512,1,1,512,0.4672
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,16,0.1498
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,32,0.1492
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,64,0.1509
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,128,0.1517
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,256,0.1519
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,512,0.1526
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,1024,0.1528
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,1,0.5474
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,2048,0.1578
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,4096,0.1656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,8192,0.1704
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1,1,1,16384,0.1798
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,4,0.5492
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,8,0.5467
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,16,0.5499
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,1,0.1517
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,4,0.1513
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,8,0.1518
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,16,0.1519
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,32,0.5517
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,32,0.1514
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,64,0.1533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,128,0.1535
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,256,0.1544
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,512,0.1547
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,1024,0.1561
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,64,0.5696
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,2048,0.1607
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,4096,0.1687
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,8192,0.1739
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,2,1,1,16384,0.1836
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,1,0.1549
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,128,0.5957
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,4,0.1545
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,8,0.1544
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,16,0.1550
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,32,0.1548
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,64,0.1563
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,128,0.1569
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,256,0.1571
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,32,1024,1,1,256,0.6554
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,512,0.1581
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,1024,0.1599
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,2048,0.1664
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,4096,0.1736
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,8192,0.1787
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,4,1,1,16384,0.1894
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,1,0.1626
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,4,0.1625
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,8,0.1628
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,16,0.1626
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,32,0.1629
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,64,0.1645
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,128,0.1656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,256,0.1661
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,512,0.1677
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,1024,0.1708
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,2048,0.1818
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,4096,0.1856
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,8192,0.1922
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,8,1,1,16384,0.2033
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,1,0.1779
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,4,0.1782
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,8,0.1777
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,16,0.1778
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,32,0.1783
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,64,0.1808
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,128,0.1819
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,256,0.1828
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,512,0.1857
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,1024,0.1891
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,2048,0.2050
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,4096,0.2101
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,8192,0.2179
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,16,1,1,16384,0.2334
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,1,0.1734
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,4,0.1732
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,8,0.1736
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,16,0.1734
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,32,0.1738
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,64,0.1764
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,128,0.1781
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,256,0.1809
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,512,0.1848
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,1024,0.1929
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,2048,0.2148
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,32,1,1,4096,0.2213
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,1,0.1794
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,4,0.1806
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,8,0.1807
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,16,0.1807
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,32,0.1808
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,64,0.1837
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,128,0.1870
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,256,0.1924
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,512,0.2010
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,1024,0.2138
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,2048,0.2512
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,64,1,1,4096,0.2617
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,1,0.1900
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,4,0.1900
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,8,0.1907
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,16,0.1904
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,32,0.1912
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,64,0.1945
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,128,0.1988
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,256,0.2061
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,512,0.2210
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,1024,0.2484
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,128,1,1,2048,0.3123
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,1,0.2171
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,4,0.2172
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,8,0.2174
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,16,0.2179
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,32,0.2185
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,64,0.2239
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,128,0.2322
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,256,0.2480
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,512,0.2777
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,256,1,1,1024,0.3329
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,1,0.2794
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,4,0.2792
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,8,0.2796
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,16,0.2795
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,32,0.2813
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,64,0.2926
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,128,0.3088
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,256,0.3408
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,512,1,1,512,0.3990
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,1,0.4150
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,4,0.4175
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,8,0.4185
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,16,0.4201
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,32,0.4232
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,64,0.4434
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,128,0.4698
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,1,1.4901
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,bfloat16,bfloat16,16,1024,1,1,256,0.5277
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,1,0.2329
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,4,1.4901
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,4,0.2333
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,8,0.2328
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,16,0.2332
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,32,0.2338
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,8,1.4903
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,64,0.2341
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,128,0.2342
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,256,0.2359
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,512,0.2377
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,1024,0.2419
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,2048,0.2535
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,16,1.4845
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1024,1,1,1,1.7421
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,4096,0.2600
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,8192,0.2653
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1,1,1,16384,0.2748
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,32,1.4861
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1024,1,1,4,1.7431
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,64,1.4914
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1024,1,1,8,1.7323
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,1,0.2370
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,4,0.2368
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,8,0.2368
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,16,0.2366
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1024,1,1,16,1.7338
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,32,0.2368
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,64,0.2371
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,128,0.2381
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,256,0.2390
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,512,0.2400
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,1024,0.2428
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1024,1,1,32,1.7355
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,2048,0.2510
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,4096,0.2579
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,128,1.5010
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,8192,0.2626
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,2,1,1,16384,0.2728
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,1,0.2425
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1024,1,1,64,1.7409
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,4,0.2419
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,8,0.2423
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,16,0.2421
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,256,1.5360
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,32,0.2428
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,64,0.2430
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,128,0.2429
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,256,0.2449
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1024,1,1,128,1.7528
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,512,0.2456
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,1024,0.2468
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,2048,0.2540
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,4096,0.2607
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,8192,0.2662
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,4,1,1,16384,0.2771
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,128,1024,1,1,256,1.7791
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,1,0.2565
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,4,0.2565
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,8,0.2569
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,16,0.2572
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,32,0.2568
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,64,0.2572
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,128,0.2577
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,256,0.2587
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,512,0.2600
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,1024,0.2619
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,2048,0.2710
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,4096,0.2762
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,8192,0.2820
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,8,1,1,16384,0.2945
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,1,0.2834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,4,0.2831
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,8,0.2828
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,16,0.2827
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,32,0.2822
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,64,0.2834
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,128,0.2838
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,256,0.2850
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,512,0.2874
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,1024,0.2880
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,2048,0.2985
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,4096,0.3032
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,8192,0.3111
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,16,1,1,16384,0.3261
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,1,0.2977
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,4,0.2972
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,8,0.2979
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,16,0.2981
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,32,0.2970
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,64,0.2980
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,128,0.2986
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,256,0.2990
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,512,0.3027
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,1024,0.3061
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,2048,0.3177
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,32,1,1,4096,0.3243
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,1,0.3287
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,4,0.3289
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,8,0.3290
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,16,0.3283
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,32,0.3286
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,64,0.3294
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,128,0.3299
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,256,0.3310
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,512,0.3334
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,1024,0.3384
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,2048,0.3564
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,64,1,1,4096,0.3665
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,1,0.4151
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,4,0.4155
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,8,0.4157
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,16,0.4158
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,32,0.4144
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,64,0.4160
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,128,0.4179
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,256,0.4190
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,512,0.4246
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,1024,0.4352
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,128,1,1,2048,0.4669
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,1,0.5792
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,4,0.5795
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,1,1.3600
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,8,0.5804
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,4,1.3455
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,16,0.5772
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,32,0.5785
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,8,1.3460
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,64,0.5820
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,128,0.5901
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,256,0.5974
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,512,0.6018
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,16,1.3451
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,256,1,1,1024,0.6262
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,32,1.3610
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,1,0.9832
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,4,0.9831
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,64,1.3656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,8,0.9800
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,16,0.9805
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,128,1.3652
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,32,0.9960
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,64,0.9978
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,deepseek-ai/DeepSeek-V3.2,DeepseekV32ForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,256,1.3926
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,128,0.9951
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,256,1.0090
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,512,1,1,512,1.0404
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,1,1.8303
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,4,1.8379
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,8,1.8317
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,16,1.8408
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,32,1.8421
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,64,1.8491
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,128,1.8635
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,64,1024,1,1,256,1.8760
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,1,0.1773
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,4,0.1775
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,8,0.1772
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,16,0.1778
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,32,0.1783
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,64,0.1785
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,128,0.1792
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,256,0.1803
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,512,0.1829
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,1024,0.1869
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,2048,0.1974
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,4096,0.2048
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,8192,0.2087
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1,1,1,16384,0.2185
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,1,0.1806
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,4,0.1807
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,8,0.1816
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,16,0.1810
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,32,0.1809
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,64,0.1811
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,128,0.1826
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,256,0.1828
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,512,0.1838
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,1024,0.1860
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,2048,0.1971
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,4096,0.2004
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,8192,0.2064
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,2,1,1,16384,0.2154
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,1,0.1856
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,4,0.1854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,8,0.1854
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,16,0.1859
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,32,0.1857
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,64,0.1859
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,128,0.1870
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,256,0.1877
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,512,0.1890
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,1024,0.1900
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,2048,0.1971
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,4096,0.2037
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,8192,0.2091
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,4,1,1,16384,0.2200
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,1,0.1983
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,4,0.1983
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,8,0.1990
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,16,0.1992
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,32,0.1992
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,64,0.1984
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,128,0.1996
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,256,0.2012
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,512,0.2016
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,1024,0.2035
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,2048,0.2127
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,4096,0.2167
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,8192,0.2229
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,8,1,1,16384,0.2363
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,1,0.2234
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,4,0.2237
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,8,0.2228
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,16,0.2227
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,32,0.2233
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,1,0.1529
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,4,0.1533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,64,0.2234
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,8,0.1534
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,128,0.2240
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,256,0.2250
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,16,0.1532
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,32,0.1533
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,512,0.2277
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,64,0.1540
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,1024,0.2282
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,128,0.1545
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,2048,0.2384
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,256,0.1560
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,4096,0.2433
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,512,0.1583
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,8192,0.2510
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,1024,0.1620
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,2048,0.1726
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,16,1,1,16384,0.2669
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,4096,0.1798
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,8192,0.1847
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,1,0.2358
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1,1,1,16384,0.1939
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,4,0.2358
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,8,0.2364
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,16,0.2369
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,1,0.1560
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,32,0.2362
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,4,0.1554
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,64,0.2365
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,8,0.1554
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,128,0.2376
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,16,0.1554
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,32,0.1559
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,256,0.2384
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,64,0.1560
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,512,0.2404
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,128,0.1570
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,256,0.1584
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,1024,0.2468
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,512,0.1584
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,1024,0.1612
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,2048,0.2563
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,2048,0.1684
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,32,1,1,4096,0.2630
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,4096,0.1753
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,8192,0.1806
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,1,0.2653
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,2,1,1,16384,0.1903
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,1,0.1594
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,4,0.2656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,4,0.1596
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,8,0.2655
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,8,0.1595
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,16,0.1598
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,16,0.2657
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,32,0.1599
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,64,0.1602
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,32,0.2654
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,128,0.1606
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,256,0.1622
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,64,0.2656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,512,0.1630
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,128,0.2667
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,1024,0.1642
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,2048,0.1708
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,256,0.2680
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,4096,0.1781
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,512,0.2713
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,8192,0.1831
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,1024,0.2752
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,4,1,1,16384,0.1939
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,1,0.1725
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,2048,0.2946
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,4,0.1727
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,8,0.1729
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,16,0.1726
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,64,1,1,4096,0.3045
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,32,0.1728
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,64,0.1727
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,128,0.1735
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,1,0.3442
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,256,0.1746
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,512,0.1753
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,4,0.3445
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,1024,0.1771
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,2048,0.1863
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,8,0.3442
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,4096,0.1901
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,16,0.3453
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,8192,0.1969
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,32,0.3445
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,8,1,1,16384,0.2098
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,64,0.3458
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,1,0.1965
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,128,0.3484
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,4,0.1959
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,8,0.1960
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,16,0.1964
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,256,0.3504
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,32,0.1965
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,64,0.1966
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,512,0.3545
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,128,0.1968
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,256,0.1982
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,512,0.2007
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,1024,0.2018
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,2048,0.2116
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,1024,0.3658
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,4096,0.2163
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,8192,0.2250
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,128,1,1,2048,0.3956
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,16,1,1,16384,0.2406
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,1,0.2084
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,1,0.5021
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,4,0.2085
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,8,0.2088
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,16,0.2085
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,4,0.5020
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,32,0.2087
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,64,0.2090
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,8,0.5026
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,128,0.2093
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,256,0.2109
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,16,0.5031
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,512,0.2128
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,1024,0.2181
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,32,0.5034
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,2048,0.2288
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,32,1,1,4096,0.2357
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,64,0.5058
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,1,0.2378
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,128,0.5081
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,4,0.2381
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,8,0.2379
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,256,0.5134
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,16,0.2383
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,32,0.2379
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,512,0.5248
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,64,0.2382
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,128,0.2380
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,256,1,1,1024,0.5489
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,256,0.2399
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,512,0.2423
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,1,0.8386
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,1024,0.2472
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,2048,0.2664
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,4,0.8388
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,64,1,1,4096,0.2766
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,1,0.3122
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,8,0.8394
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,4,0.3118
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,8,0.3118
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,16,0.8396
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,16,0.3123
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,32,0.3129
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,32,0.8416
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,64,0.3138
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,128,0.3152
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,64,0.8438
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,256,0.3186
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,512,0.3230
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,128,0.8504
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,1024,0.3338
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,256,0.8647
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,128,1,1,2048,0.3649
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,1,0.4652
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,512,1,1,512,0.8893
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,4,0.4666
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,8,0.4656
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,1,1.5388
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,16,0.4668
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,32,0.4693
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,4,1.5401
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,64,0.4693
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,128,0.4717
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,8,1.5308
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,256,0.4771
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,512,0.4890
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,16,1.5433
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,256,1,1,1024,0.5115
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,32,1.5447
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,1,0.7749
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,64,1.5489
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,4,0.7737
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,8,0.7751
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,128,1.5605
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,16,0.7781
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,32,0.7749
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,32,1024,1,1,256,1.5866
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,64,0.7772
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,128,0.7855
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,256,0.7994
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,512,1,1,512,0.8210
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,1,1.4075
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,4,1.3915
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,8,1.4061
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,16,1.3972
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,32,1.3979
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,64,1.4124
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,128,1.4165
SGLang,0.5.9,NVIDIA H100 80GB HBM3,dsa_generation_module,dsa_nsa,zai-org/GLM-5,GlmMoeDsaForCausalLM,bfloat16,fp8,bfloat16,16,1024,1,1,256,1.4606
