[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-qwen3-6-35b-a3b-fp8":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":17,"family":26,"score":28,"variants":29,"quantizations":30,"requirements":57,"gpu_recommendations":133,"fitting_gpus":196,"community":227},458,"qwen3-6-35b-a3b-fp8","Qwen\u002FQwen3.6-35B-A3B-FP8","Qwen","qwen3_5_moe",{"layers":10,"head_dim":11,"kv_heads":12,"hidden_size":13},40,256,2,2048,2904031232,"2.9B",262144,null,11088336,351,"published","https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen3.6-35B-A3B-FP8","2026-08-13T18:03:27+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":7,"slug":27},"qwen",0.0225,[],[31,36,41,46,50,54],{"format":32,"bits":33,"size_factor":34,"quality_loss":35},"AWQ",4,0.263,0.03,{"format":37,"bits":38,"size_factor":39,"quality_loss":40},"FP16",16,1,0,{"format":42,"bits":43,"size_factor":44,"quality_loss":45},"FP8",8,0.5,0.01,{"format":47,"bits":33,"size_factor":48,"quality_loss":49},"GGUF",0.25,0.05,{"format":51,"bits":33,"size_factor":52,"quality_loss":53},"GPTQ",0.275,0.04,{"format":55,"bits":43,"size_factor":44,"quality_loss":56},"INT8",0.02,{"FP16":58,"FP8":81,"INT8":93,"AWQ":97,"GPTQ":109,"GGUF":121},{"minimum":59,"production":69,"recommended":75},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":62,"kv_cache_gb":63,"vram_gb":64,"ram_gb":65,"disk_gb":66,"requirement_source":67,"confidence":68},"minimum",4096,5.41,0.31,10.76,32,47.44,"inferred",0.95,{"scenario":70,"context_length":71,"batch_size":33,"weight_gb":62,"kv_cache_gb":10,"vram_gb":72,"ram_gb":73,"disk_gb":74,"requirement_source":67,"confidence":68},"production",16384,81.86,122.8,112.44,{"scenario":76,"context_length":77,"batch_size":12,"weight_gb":62,"kv_cache_gb":78,"vram_gb":79,"ram_gb":65,"disk_gb":80,"requirement_source":67,"confidence":68},"recommended",8192,5,20.95,73.44,{"minimum":82,"production":86,"recommended":90},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":83,"kv_cache_gb":63,"vram_gb":84,"ram_gb":65,"disk_gb":85,"requirement_source":67,"confidence":68},2.7,7.34,43.22,{"scenario":70,"context_length":71,"batch_size":33,"weight_gb":83,"kv_cache_gb":10,"vram_gb":87,"ram_gb":88,"disk_gb":89,"requirement_source":67,"confidence":68},78.13,117.2,108.22,{"scenario":76,"context_length":77,"batch_size":12,"weight_gb":83,"kv_cache_gb":78,"vram_gb":91,"ram_gb":65,"disk_gb":92,"requirement_source":67,"confidence":68},17.38,69.22,{"minimum":94,"production":95,"recommended":96},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":83,"kv_cache_gb":63,"vram_gb":84,"ram_gb":65,"disk_gb":85,"requirement_source":67,"confidence":68},{"scenario":70,"context_length":71,"batch_size":33,"weight_gb":83,"kv_cache_gb":10,"vram_gb":87,"ram_gb":88,"disk_gb":89,"requirement_source":67,"confidence":68},{"scenario":76,"context_length":77,"batch_size":12,"weight_gb":83,"kv_cache_gb":78,"vram_gb":91,"ram_gb":65,"disk_gb":92,"requirement_source":67,"confidence":68},{"minimum":98,"production":102,"recommended":106},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":99,"kv_cache_gb":63,"vram_gb":100,"ram_gb":65,"disk_gb":101,"requirement_source":67,"confidence":68},1.4,5.69,41.19,{"scenario":70,"context_length":71,"batch_size":33,"weight_gb":99,"kv_cache_gb":10,"vram_gb":103,"ram_gb":104,"disk_gb":105,"requirement_source":67,"confidence":68},76.34,114.5,106.19,{"scenario":76,"context_length":77,"batch_size":12,"weight_gb":99,"kv_cache_gb":78,"vram_gb":107,"ram_gb":65,"disk_gb":108,"requirement_source":67,"confidence":68},15.66,67.19,{"minimum":110,"production":114,"recommended":118},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":111,"kv_cache_gb":63,"vram_gb":112,"ram_gb":65,"disk_gb":113,"requirement_source":67,"confidence":68},1.42,5.71,41.22,{"scenario":70,"context_length":71,"batch_size":33,"weight_gb":111,"kv_cache_gb":10,"vram_gb":115,"ram_gb":116,"disk_gb":117,"requirement_source":67,"confidence":68},76.36,114.54,106.22,{"scenario":76,"context_length":77,"batch_size":12,"weight_gb":111,"kv_cache_gb":78,"vram_gb":119,"ram_gb":65,"disk_gb":120,"requirement_source":67,"confidence":68},15.68,67.22,{"minimum":122,"production":126,"recommended":130},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":123,"kv_cache_gb":63,"vram_gb":124,"ram_gb":65,"disk_gb":125,"requirement_source":67,"confidence":68},1.45,5.76,41.27,{"scenario":70,"context_length":71,"batch_size":33,"weight_gb":123,"kv_cache_gb":10,"vram_gb":127,"ram_gb":128,"disk_gb":129,"requirement_source":67,"confidence":68},76.41,114.61,106.27,{"scenario":76,"context_length":77,"batch_size":12,"weight_gb":123,"kv_cache_gb":78,"vram_gb":131,"ram_gb":65,"disk_gb":132,"requirement_source":67,"confidence":68},15.72,67.27,{"recommended":134,"cheapest":151,"performance":160,"min_complexity":177,"cluster_options":192,"no_match":193,"required_vram_gb":107,"candidate_count":194,"usage":195},{"gpu_model":135,"gpu_model_id":12,"gpu_count":39,"total_vram_gb":65,"hour_price":136,"monthly_cost":137,"providers":138,"bandwidth_gbps":141,"fp16_tflops":142,"interconnect_score":10,"parallel_efficiency":39,"score":143,"tier":144,"perf_metric":145,"perf_label":146,"perf_value":147,"workload":17,"cost_metric":148},"RTX 5090",0.3337,243.6,[139,140],"vast","runpod",1792,209,0.6282,"cheapest","tok\u002Fs","1280 tok\u002Fs",1280,{"effective_value":149,"effective_unit":145,"capacity_factor":150},896,0.7,{"gpu_model":152,"gpu_model_id":153,"gpu_count":39,"total_vram_gb":38,"hour_price":154,"monthly_cost":155,"providers":156,"bandwidth_gbps":40,"fp16_tflops":40,"interconnect_score":10,"parallel_efficiency":39,"score":157,"tier":144,"perf_metric":145,"perf_label":158,"perf_value":40,"workload":17,"cost_metric":159},"Tesla V100",17,0.0289,21.1,[139],0.6362,"0 tok\u002Fs",{"effective_value":40,"effective_unit":145,"capacity_factor":150},{"gpu_model":161,"gpu_model_id":162,"gpu_count":43,"total_vram_gb":163,"hour_price":164,"monthly_cost":165,"providers":166,"bandwidth_gbps":168,"fp16_tflops":169,"interconnect_score":10,"parallel_efficiency":170,"score":171,"tier":172,"perf_metric":145,"perf_label":173,"perf_value":174,"workload":17,"cost_metric":175},"H200 SXM",11,1128,3.5,20440,[167],"lambda",4800,989,0.4,0.2197,"balanced","10971.4 tok\u002Fs",10971.4,{"effective_value":176,"effective_unit":145,"capacity_factor":150},7680,{"gpu_model":178,"gpu_model_id":179,"gpu_count":39,"total_vram_gb":180,"hour_price":181,"monthly_cost":182,"providers":183,"bandwidth_gbps":185,"fp16_tflops":186,"interconnect_score":10,"parallel_efficiency":39,"score":187,"tier":144,"perf_metric":145,"perf_label":188,"perf_value":189,"workload":17,"cost_metric":190},"RTX 3090",3,24,0.0622,45.41,[139,184],"tensordock",936,71,0.6697,"668.6 tok\u002Fs",668.6,{"effective_value":191,"effective_unit":145,"capacity_factor":150},468,[],false,26,"value",[197,201,205,212,219],{"name":152,"vram_gb":38,"cheapest_hour":154,"cost":198,"provider":139,"estimated_tps":40},{"hourly":154,"daily":199,"monthly":155,"yearly":200},0.69,253.16,{"name":178,"vram_gb":180,"cheapest_hour":181,"cost":202,"provider":139,"estimated_tps":189},{"hourly":181,"daily":203,"monthly":182,"yearly":204},1.49,544.87,{"name":206,"vram_gb":38,"cheapest_hour":207,"cost":208,"provider":139,"estimated_tps":40},"RTX 4070S Ti",0.0678,{"hourly":207,"daily":209,"monthly":210,"yearly":211},1.63,49.49,593.93,{"name":213,"vram_gb":38,"cheapest_hour":214,"cost":215,"provider":139,"estimated_tps":40},"RTX 4080S",0.0685,{"hourly":214,"daily":216,"monthly":217,"yearly":218},1.64,50.01,600.06,{"name":220,"vram_gb":38,"cheapest_hour":221,"cost":222,"provider":139,"estimated_tps":226},"RTX 5070 Ti",0.0849,{"hourly":221,"daily":223,"monthly":224,"yearly":225},2.04,61.98,743.72,640,{"posts_count":40,"benchmarks_count":40}]