[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-qwen25-vl-72b":3},{"id":4,"slug":5,"name":6,"hf_id":7,"organization":8,"architecture":9,"architecture_config":10,"parameters":11,"parameters_label":12,"context_length":13,"license":14,"downloads":15,"likes":16,"status":17,"description":10,"huggingface_url":18,"created_at":19,"seo":20,"category":23,"family":26,"score":28,"variants":29,"quantizations":30,"requirements":57,"gpu_recommendations":141,"fitting_gpus":211,"community":238},320,"qwen25-vl-72b","Qwen2.5-VL 72B","Qwen\u002FQwen2.5-VL-72B-Instruct","Qwen","transformer",null,72700000000,"72.7B",32768,"MIT",528890,647,"published","https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen2.5-VL-72B-Instruct","2026-08-12T03:11:22+00:00",{"title":6,"description":21,"image":10,"robots":22,"canonical_url":10},"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":24,"slug":25},"OCR \u002F Vision","ocr-vision",{"name":8,"slug":27},"qwen",0.008,[],[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":95,"AWQ":99,"GPTQ":113,"GGUF":127},{"minimum":59,"production":68,"recommended":75},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":62,"kv_cache_gb":48,"vram_gb":63,"ram_gb":64,"disk_gb":65,"requirement_source":66,"confidence":67},"minimum",2048,135.41,174.87,262.31,250.25,"estimated",0.95,{"scenario":69,"context_length":70,"batch_size":71,"weight_gb":62,"kv_cache_gb":43,"vram_gb":72,"ram_gb":73,"disk_gb":74,"requirement_source":66,"confidence":67},"production",8192,2,208.47,312.71,315.25,{"scenario":76,"context_length":77,"batch_size":39,"weight_gb":62,"kv_cache_gb":39,"vram_gb":78,"ram_gb":79,"disk_gb":80,"requirement_source":66,"confidence":67},"recommended",4096,184.84,277.25,276.25,{"minimum":82,"production":87,"recommended":91},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":83,"kv_cache_gb":48,"vram_gb":84,"ram_gb":85,"disk_gb":86,"requirement_source":66,"confidence":67},67.71,89.22,133.84,144.62,{"scenario":69,"context_length":70,"batch_size":71,"weight_gb":83,"kv_cache_gb":43,"vram_gb":88,"ram_gb":89,"disk_gb":90,"requirement_source":66,"confidence":67},115.04,172.55,209.62,{"scenario":76,"context_length":77,"batch_size":39,"weight_gb":83,"kv_cache_gb":39,"vram_gb":92,"ram_gb":93,"disk_gb":94,"requirement_source":66,"confidence":67},95.29,142.94,170.62,{"minimum":96,"production":97,"recommended":98},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":83,"kv_cache_gb":48,"vram_gb":84,"ram_gb":85,"disk_gb":86,"requirement_source":66,"confidence":67},{"scenario":69,"context_length":70,"batch_size":71,"weight_gb":83,"kv_cache_gb":43,"vram_gb":88,"ram_gb":89,"disk_gb":90,"requirement_source":66,"confidence":67},{"scenario":76,"context_length":77,"batch_size":39,"weight_gb":83,"kv_cache_gb":39,"vram_gb":92,"ram_gb":93,"disk_gb":94,"requirement_source":66,"confidence":67},{"minimum":100,"production":105,"recommended":109},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":101,"kv_cache_gb":48,"vram_gb":102,"ram_gb":103,"disk_gb":104,"requirement_source":66,"confidence":67},35.12,48.01,72.01,93.79,{"scenario":69,"context_length":70,"batch_size":71,"weight_gb":101,"kv_cache_gb":43,"vram_gb":106,"ram_gb":107,"disk_gb":108,"requirement_source":66,"confidence":67},70.07,105.1,158.79,{"scenario":76,"context_length":77,"batch_size":39,"weight_gb":101,"kv_cache_gb":39,"vram_gb":110,"ram_gb":111,"disk_gb":112,"requirement_source":66,"confidence":67},52.2,78.3,119.79,{"minimum":114,"production":119,"recommended":123},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":115,"kv_cache_gb":48,"vram_gb":116,"ram_gb":117,"disk_gb":118,"requirement_source":66,"confidence":67},35.55,48.54,72.81,94.45,{"scenario":69,"context_length":70,"batch_size":71,"weight_gb":115,"kv_cache_gb":43,"vram_gb":120,"ram_gb":121,"disk_gb":122,"requirement_source":66,"confidence":67},70.65,105.98,159.45,{"scenario":76,"context_length":77,"batch_size":39,"weight_gb":115,"kv_cache_gb":39,"vram_gb":124,"ram_gb":125,"disk_gb":126,"requirement_source":66,"confidence":67},52.76,79.14,120.45,{"minimum":128,"production":133,"recommended":137},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":129,"kv_cache_gb":48,"vram_gb":130,"ram_gb":131,"disk_gb":132,"requirement_source":66,"confidence":67},36.39,49.61,74.42,95.77,{"scenario":69,"context_length":70,"batch_size":71,"weight_gb":129,"kv_cache_gb":43,"vram_gb":134,"ram_gb":135,"disk_gb":136,"requirement_source":66,"confidence":67},71.82,107.73,160.77,{"scenario":76,"context_length":77,"batch_size":39,"weight_gb":129,"kv_cache_gb":39,"vram_gb":138,"ram_gb":139,"disk_gb":140,"requirement_source":66,"confidence":67},53.88,80.82,121.77,{"recommended":142,"cheapest":164,"performance":180,"min_complexity":194,"cluster_options":207,"no_match":208,"required_vram_gb":110,"candidate_count":209,"usage":210},{"gpu_model":143,"gpu_model_id":144,"gpu_count":39,"total_vram_gb":145,"hour_price":146,"monthly_cost":147,"providers":148,"bandwidth_gbps":153,"fp16_tflops":154,"interconnect_score":155,"parallel_efficiency":39,"score":156,"tier":157,"perf_metric":158,"perf_label":159,"perf_value":160,"workload":10,"cost_metric":161},"H100 SXM",10,80,1.6059,1172.31,[149,150,151,152],"vast","tensordock","runpod","lambda",3350,66.9,40,0.6635,"balanced","tok\u002Fs","95.4 tok\u002Fs",95.4,{"effective_value":162,"effective_unit":158,"capacity_factor":163},66.8,0.7,{"gpu_model":165,"gpu_model_id":166,"gpu_count":43,"total_vram_gb":167,"hour_price":168,"monthly_cost":169,"providers":170,"bandwidth_gbps":171,"fp16_tflops":172,"interconnect_score":155,"parallel_efficiency":173,"score":174,"tier":175,"perf_metric":158,"perf_label":176,"perf_value":177,"workload":10,"cost_metric":178},"RTX 5070",28,96,0.0101,58.98,[149],672,61,0.4,0.5191,"cheapest","61.2 tok\u002Fs",61.2,{"effective_value":179,"effective_unit":158,"capacity_factor":163},42.8,{"gpu_model":181,"gpu_model_id":182,"gpu_count":43,"total_vram_gb":183,"hour_price":184,"monthly_cost":185,"providers":186,"bandwidth_gbps":187,"fp16_tflops":188,"interconnect_score":155,"parallel_efficiency":173,"score":189,"tier":157,"perf_metric":158,"perf_label":190,"perf_value":191,"workload":10,"cost_metric":192},"H200 SXM",11,1128,3.5,20440,[152],4800,989,0.2197,"437.4 tok\u002Fs",437.4,{"effective_value":193,"effective_unit":158,"capacity_factor":163},306.2,{"gpu_model":195,"gpu_model_id":196,"gpu_count":39,"total_vram_gb":145,"hour_price":197,"monthly_cost":198,"providers":199,"bandwidth_gbps":200,"fp16_tflops":201,"interconnect_score":155,"parallel_efficiency":39,"score":202,"tier":157,"perf_metric":158,"perf_label":203,"perf_value":204,"workload":10,"cost_metric":205},"A100 80GB",9,1.2,876,[149,150,151,152],2039,77.97,0.6603,"58.1 tok\u002Fs",58.1,{"effective_value":206,"effective_unit":158,"capacity_factor":163},40.7,[],false,19,"value",[212,216,220,227,235],{"name":195,"vram_gb":145,"cheapest_hour":197,"cost":213,"provider":149,"estimated_tps":204},{"hourly":197,"daily":214,"monthly":198,"yearly":215},28.8,10512,{"name":143,"vram_gb":145,"cheapest_hour":146,"cost":217,"provider":149,"estimated_tps":160},{"hourly":146,"daily":218,"monthly":147,"yearly":219},38.54,14067.68,{"name":181,"vram_gb":221,"cheapest_hour":184,"cost":222,"provider":152,"estimated_tps":226},141,{"hourly":184,"daily":223,"monthly":224,"yearly":225},84,2555,30660,136.7,{"name":228,"vram_gb":229,"cheapest_hour":230,"cost":231,"provider":149,"estimated_tps":40},"B200",180,4.1271,{"hourly":230,"daily":232,"monthly":233,"yearly":234},99.05,3012.78,36153.4,{"name":236,"vram_gb":229,"cheapest_hour":10,"cost":10,"provider":10,"estimated_tps":237},"B200 SXM",222.1,{"posts_count":40,"benchmarks_count":40}]