[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-qwen3-6-35b-a3b-nvfp4":3},{"id":4,"slug":5,"name":6,"hf_id":6,"organization":7,"architecture":8,"architecture_config":9,"parameters":14,"parameters_label":15,"context_length":16,"license":17,"downloads":18,"likes":19,"status":20,"description":17,"huggingface_url":21,"created_at":22,"seo":23,"category":26,"family":29,"score":32,"variants":33,"quantizations":34,"requirements":61,"gpu_recommendations":137,"fitting_gpus":200,"community":231},455,"qwen3-6-35b-a3b-nvfp4","nvidia\u002FQwen3.6-35B-A3B-NVFP4","nvidia","qwen3_5_moe",{"layers":10,"head_dim":11,"kv_heads":12,"hidden_size":13},40,256,2,2048,2904031232,"2.9B",262144,null,11698237,552,"published","https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002FQwen3.6-35B-A3B-NVFP4","2026-08-13T18:03:12+00:00",{"title":6,"description":24,"image":17,"robots":25,"canonical_url":17},"AI 模型部署决策引擎——开源 AI 的 GPU 需求、云端价格与成本分析。","index, follow",{"name":27,"slug":28},"LLM","llm",{"name":30,"slug":31},"Qwen","qwen",0.0245,[],[35,40,45,50,54,58],{"format":36,"bits":37,"size_factor":38,"quality_loss":39},"AWQ",4,0.263,0.03,{"format":41,"bits":42,"size_factor":43,"quality_loss":44},"FP16",16,1,0,{"format":46,"bits":47,"size_factor":48,"quality_loss":49},"FP8",8,0.5,0.01,{"format":51,"bits":37,"size_factor":52,"quality_loss":53},"GGUF",0.25,0.05,{"format":55,"bits":37,"size_factor":56,"quality_loss":57},"GPTQ",0.275,0.04,{"format":59,"bits":47,"size_factor":48,"quality_loss":60},"INT8",0.02,{"FP16":62,"FP8":85,"INT8":97,"AWQ":101,"GPTQ":113,"GGUF":125},{"minimum":63,"production":73,"recommended":79},{"scenario":64,"context_length":65,"batch_size":43,"weight_gb":66,"kv_cache_gb":67,"vram_gb":68,"ram_gb":69,"disk_gb":70,"requirement_source":71,"confidence":72},"minimum",4096,5.41,0.31,10.76,32,47.44,"inferred",0.95,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":66,"kv_cache_gb":10,"vram_gb":76,"ram_gb":77,"disk_gb":78,"requirement_source":71,"confidence":72},"production",16384,81.86,122.8,112.44,{"scenario":80,"context_length":81,"batch_size":12,"weight_gb":66,"kv_cache_gb":82,"vram_gb":83,"ram_gb":69,"disk_gb":84,"requirement_source":71,"confidence":72},"recommended",8192,5,20.95,73.44,{"minimum":86,"production":90,"recommended":94},{"scenario":64,"context_length":65,"batch_size":43,"weight_gb":87,"kv_cache_gb":67,"vram_gb":88,"ram_gb":69,"disk_gb":89,"requirement_source":71,"confidence":72},2.7,7.34,43.22,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":87,"kv_cache_gb":10,"vram_gb":91,"ram_gb":92,"disk_gb":93,"requirement_source":71,"confidence":72},78.13,117.2,108.22,{"scenario":80,"context_length":81,"batch_size":12,"weight_gb":87,"kv_cache_gb":82,"vram_gb":95,"ram_gb":69,"disk_gb":96,"requirement_source":71,"confidence":72},17.38,69.22,{"minimum":98,"production":99,"recommended":100},{"scenario":64,"context_length":65,"batch_size":43,"weight_gb":87,"kv_cache_gb":67,"vram_gb":88,"ram_gb":69,"disk_gb":89,"requirement_source":71,"confidence":72},{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":87,"kv_cache_gb":10,"vram_gb":91,"ram_gb":92,"disk_gb":93,"requirement_source":71,"confidence":72},{"scenario":80,"context_length":81,"batch_size":12,"weight_gb":87,"kv_cache_gb":82,"vram_gb":95,"ram_gb":69,"disk_gb":96,"requirement_source":71,"confidence":72},{"minimum":102,"production":106,"recommended":110},{"scenario":64,"context_length":65,"batch_size":43,"weight_gb":103,"kv_cache_gb":67,"vram_gb":104,"ram_gb":69,"disk_gb":105,"requirement_source":71,"confidence":72},1.4,5.69,41.19,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":103,"kv_cache_gb":10,"vram_gb":107,"ram_gb":108,"disk_gb":109,"requirement_source":71,"confidence":72},76.34,114.5,106.19,{"scenario":80,"context_length":81,"batch_size":12,"weight_gb":103,"kv_cache_gb":82,"vram_gb":111,"ram_gb":69,"disk_gb":112,"requirement_source":71,"confidence":72},15.66,67.19,{"minimum":114,"production":118,"recommended":122},{"scenario":64,"context_length":65,"batch_size":43,"weight_gb":115,"kv_cache_gb":67,"vram_gb":116,"ram_gb":69,"disk_gb":117,"requirement_source":71,"confidence":72},1.42,5.71,41.22,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":115,"kv_cache_gb":10,"vram_gb":119,"ram_gb":120,"disk_gb":121,"requirement_source":71,"confidence":72},76.36,114.54,106.22,{"scenario":80,"context_length":81,"batch_size":12,"weight_gb":115,"kv_cache_gb":82,"vram_gb":123,"ram_gb":69,"disk_gb":124,"requirement_source":71,"confidence":72},15.68,67.22,{"minimum":126,"production":130,"recommended":134},{"scenario":64,"context_length":65,"batch_size":43,"weight_gb":127,"kv_cache_gb":67,"vram_gb":128,"ram_gb":69,"disk_gb":129,"requirement_source":71,"confidence":72},1.45,5.76,41.27,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":127,"kv_cache_gb":10,"vram_gb":131,"ram_gb":132,"disk_gb":133,"requirement_source":71,"confidence":72},76.41,114.61,106.27,{"scenario":80,"context_length":81,"batch_size":12,"weight_gb":127,"kv_cache_gb":82,"vram_gb":135,"ram_gb":69,"disk_gb":136,"requirement_source":71,"confidence":72},15.72,67.27,{"recommended":138,"cheapest":155,"performance":164,"min_complexity":181,"cluster_options":196,"no_match":197,"required_vram_gb":111,"candidate_count":198,"usage":199},{"gpu_model":139,"gpu_model_id":12,"gpu_count":43,"total_vram_gb":69,"hour_price":140,"monthly_cost":141,"providers":142,"bandwidth_gbps":145,"fp16_tflops":146,"interconnect_score":10,"parallel_efficiency":43,"score":147,"tier":148,"perf_metric":149,"perf_label":150,"perf_value":151,"workload":17,"cost_metric":152},"RTX 5090",0.3337,243.6,[143,144],"vast","runpod",1792,209,0.6282,"cheapest","tok\u002Fs","1280 tok\u002Fs",1280,{"effective_value":153,"effective_unit":149,"capacity_factor":154},896,0.7,{"gpu_model":156,"gpu_model_id":157,"gpu_count":43,"total_vram_gb":42,"hour_price":158,"monthly_cost":159,"providers":160,"bandwidth_gbps":44,"fp16_tflops":44,"interconnect_score":10,"parallel_efficiency":43,"score":161,"tier":148,"perf_metric":149,"perf_label":162,"perf_value":44,"workload":17,"cost_metric":163},"Tesla V100",17,0.0289,21.1,[143],0.6362,"0 tok\u002Fs",{"effective_value":44,"effective_unit":149,"capacity_factor":154},{"gpu_model":165,"gpu_model_id":166,"gpu_count":47,"total_vram_gb":167,"hour_price":168,"monthly_cost":169,"providers":170,"bandwidth_gbps":172,"fp16_tflops":173,"interconnect_score":10,"parallel_efficiency":174,"score":175,"tier":176,"perf_metric":149,"perf_label":177,"perf_value":178,"workload":17,"cost_metric":179},"H200 SXM",11,1128,3.5,20440,[171],"lambda",4800,989,0.4,0.2197,"balanced","10971.4 tok\u002Fs",10971.4,{"effective_value":180,"effective_unit":149,"capacity_factor":154},7680,{"gpu_model":182,"gpu_model_id":183,"gpu_count":43,"total_vram_gb":184,"hour_price":185,"monthly_cost":186,"providers":187,"bandwidth_gbps":189,"fp16_tflops":190,"interconnect_score":10,"parallel_efficiency":43,"score":191,"tier":148,"perf_metric":149,"perf_label":192,"perf_value":193,"workload":17,"cost_metric":194},"RTX 3090",3,24,0.0622,45.41,[143,188],"tensordock",936,71,0.6697,"668.6 tok\u002Fs",668.6,{"effective_value":195,"effective_unit":149,"capacity_factor":154},468,[],false,26,"value",[201,205,209,216,223],{"name":156,"vram_gb":42,"cheapest_hour":158,"cost":202,"provider":143,"estimated_tps":44},{"hourly":158,"daily":203,"monthly":159,"yearly":204},0.69,253.16,{"name":182,"vram_gb":184,"cheapest_hour":185,"cost":206,"provider":143,"estimated_tps":193},{"hourly":185,"daily":207,"monthly":186,"yearly":208},1.49,544.87,{"name":210,"vram_gb":42,"cheapest_hour":211,"cost":212,"provider":143,"estimated_tps":44},"RTX 4070S Ti",0.0678,{"hourly":211,"daily":213,"monthly":214,"yearly":215},1.63,49.49,593.93,{"name":217,"vram_gb":42,"cheapest_hour":218,"cost":219,"provider":143,"estimated_tps":44},"RTX 4080S",0.0685,{"hourly":218,"daily":220,"monthly":221,"yearly":222},1.64,50.01,600.06,{"name":224,"vram_gb":42,"cheapest_hour":225,"cost":226,"provider":143,"estimated_tps":230},"RTX 5070 Ti",0.0849,{"hourly":225,"daily":227,"monthly":228,"yearly":229},2.04,61.98,743.72,640,{"posts_count":44,"benchmarks_count":44}]