framework,version,device,op_name,kernel_source,quant_type,num_token,hidden_size,intermediate_size,avg_ms
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,1,7168,2048,0.04559040069580078
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,2,7168,2048,0.04352000057697296
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,4,7168,2048,0.045110398530960084
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,8,7168,2048,0.04565440118312836
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,16,7168,2048,0.045731198787689206
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,32,7168,2048,0.04359039962291718
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,64,7168,2048,0.045398399233818054
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,128,7168,2048,0.045747199654579164
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,256,7168,2048,0.04951359927654266
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,512,7168,2048,0.059910398721694944
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,1024,7168,2048,0.08451520204544068
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,2048,7168,2048,0.12447359561920165
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,4096,7168,2048,0.2024768114089966
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,8192,7168,2048,0.3638848066329956
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,16384,7168,2048,0.6768703937530518
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,32768,7168,2048,1.2982975959777832
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,65536,7168,2048,2.5426847457885744
SGLang,0.5.8.post1,NVIDIA GB200,mlp,deepseek_v3,fp8_block,131072,7168,2048,5.017926406860352
