framework,version,device,op_name,kernel_source,quant_type,num_token,hidden_size,intermediate_size,avg_ms
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,1,7168,2048,0.04146240055561066
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,2,7168,2048,0.032924801111221313
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,4,7168,2048,0.03203200101852417
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,8,7168,2048,0.03380799889564514
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,16,7168,2048,0.03195199966430664
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,32,7168,2048,0.02974080145359039
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,64,7168,2048,0.024188800156116484
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,128,7168,2048,0.029504001140594482
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,256,7168,2048,0.038089600205421445
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,512,7168,2048,0.05373439788818359
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,1024,7168,2048,0.08829119801521301
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,2048,7168,2048,0.1611840009689331
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,4096,7168,2048,0.28758399486541747
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,8192,7168,2048,0.5591104030609131
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,16384,7168,2048,1.097324752807617
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,32768,7168,2048,2.1598943710327148
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,65536,7168,2048,4.275123214721679
SGLang,0.5.6.post2,NVIDIA H200,mlp,deepseek_v3,fp8_block,131072,7168,2048,8.933033752441407
