[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-llama-32-90b-vision":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":29,"variants":30,"quantizations":31,"requirements":58,"gpu_recommendations":142,"fitting_gpus":205,"community":233},209,"llama-32-90b-vision","Llama 3.2 90B Vision","meta-llama\u002FLlama-3.2-90B-Vision","meta-llama","transformer",null,90000000000,"90B",131072,"MIT",113,134,"published","https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FLlama-3.2-90B-Vision","2026-08-12T03:11:16+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":27,"slug":28},"Llama","llama",0.0015,[],[32,37,42,47,51,55],{"format":33,"bits":34,"size_factor":35,"quality_loss":36},"AWQ",4,0.263,0.03,{"format":38,"bits":39,"size_factor":40,"quality_loss":41},"FP16",16,1,0,{"format":43,"bits":44,"size_factor":45,"quality_loss":46},"FP8",8,0.5,0.01,{"format":48,"bits":34,"size_factor":49,"quality_loss":50},"GGUF",0.25,0.05,{"format":52,"bits":34,"size_factor":53,"quality_loss":54},"GPTQ",0.275,0.04,{"format":56,"bits":44,"size_factor":45,"quality_loss":57},"INT8",0.02,{"FP16":59,"FP8":82,"INT8":96,"AWQ":100,"GPTQ":114,"GGUF":128},{"minimum":60,"production":69,"recommended":76},{"scenario":61,"context_length":62,"batch_size":40,"weight_gb":63,"kv_cache_gb":49,"vram_gb":64,"ram_gb":65,"disk_gb":66,"requirement_source":67,"confidence":68},"minimum",2048,167.64,215.64,323.46,300.52,"estimated",0.95,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":63,"kv_cache_gb":44,"vram_gb":73,"ram_gb":74,"disk_gb":75,"requirement_source":67,"confidence":68},"production",8192,2,252.94,379.41,365.52,{"scenario":77,"context_length":78,"batch_size":40,"weight_gb":63,"kv_cache_gb":40,"vram_gb":79,"ram_gb":80,"disk_gb":81,"requirement_source":67,"confidence":68},"recommended",4096,227.45,341.18,326.52,{"minimum":83,"production":88,"recommended":92},{"scenario":61,"context_length":62,"batch_size":40,"weight_gb":84,"kv_cache_gb":49,"vram_gb":85,"ram_gb":86,"disk_gb":87,"requirement_source":67,"confidence":68},83.82,109.61,164.41,169.76,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":84,"kv_cache_gb":44,"vram_gb":89,"ram_gb":90,"disk_gb":91,"requirement_source":67,"confidence":68},137.27,205.91,234.76,{"scenario":77,"context_length":78,"batch_size":40,"weight_gb":84,"kv_cache_gb":40,"vram_gb":93,"ram_gb":94,"disk_gb":95,"requirement_source":67,"confidence":68},116.6,174.9,195.76,{"minimum":97,"production":98,"recommended":99},{"scenario":61,"context_length":62,"batch_size":40,"weight_gb":84,"kv_cache_gb":49,"vram_gb":85,"ram_gb":86,"disk_gb":87,"requirement_source":67,"confidence":68},{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":84,"kv_cache_gb":44,"vram_gb":89,"ram_gb":90,"disk_gb":91,"requirement_source":67,"confidence":68},{"scenario":77,"context_length":78,"batch_size":40,"weight_gb":84,"kv_cache_gb":40,"vram_gb":93,"ram_gb":94,"disk_gb":95,"requirement_source":67,"confidence":68},{"minimum":101,"production":106,"recommended":110},{"scenario":61,"context_length":62,"batch_size":40,"weight_gb":102,"kv_cache_gb":49,"vram_gb":103,"ram_gb":104,"disk_gb":105,"requirement_source":67,"confidence":68},43.48,58.58,87.87,106.83,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":102,"kv_cache_gb":44,"vram_gb":107,"ram_gb":108,"disk_gb":109,"requirement_source":67,"confidence":68},81.6,122.41,171.83,{"scenario":77,"context_length":78,"batch_size":40,"weight_gb":102,"kv_cache_gb":40,"vram_gb":111,"ram_gb":112,"disk_gb":113,"requirement_source":67,"confidence":68},63.25,94.88,132.83,{"minimum":115,"production":120,"recommended":124},{"scenario":61,"context_length":62,"batch_size":40,"weight_gb":116,"kv_cache_gb":49,"vram_gb":117,"ram_gb":118,"disk_gb":119,"requirement_source":67,"confidence":68},44,59.24,88.86,107.65,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":116,"kv_cache_gb":44,"vram_gb":121,"ram_gb":122,"disk_gb":123,"requirement_source":67,"confidence":68},82.33,123.49,172.65,{"scenario":77,"context_length":78,"batch_size":40,"weight_gb":116,"kv_cache_gb":40,"vram_gb":125,"ram_gb":126,"disk_gb":127,"requirement_source":67,"confidence":68},63.95,95.92,133.65,{"minimum":129,"production":134,"recommended":138},{"scenario":61,"context_length":62,"batch_size":40,"weight_gb":130,"kv_cache_gb":49,"vram_gb":131,"ram_gb":132,"disk_gb":133,"requirement_source":67,"confidence":68},45.05,60.57,90.85,109.28,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":130,"kv_cache_gb":44,"vram_gb":135,"ram_gb":136,"disk_gb":137,"requirement_source":67,"confidence":68},83.77,125.66,174.28,{"scenario":77,"context_length":78,"batch_size":40,"weight_gb":130,"kv_cache_gb":40,"vram_gb":139,"ram_gb":140,"disk_gb":141,"requirement_source":67,"confidence":68},65.33,98,135.28,{"recommended":143,"cheapest":162,"performance":175,"min_complexity":185,"cluster_options":201,"no_match":202,"required_vram_gb":111,"candidate_count":203,"usage":204},{"gpu_model":144,"gpu_model_id":145,"gpu_count":40,"total_vram_gb":146,"hour_price":147,"monthly_cost":148,"providers":149,"bandwidth_gbps":151,"fp16_tflops":152,"interconnect_score":153,"parallel_efficiency":40,"score":154,"tier":155,"perf_metric":156,"perf_label":157,"perf_value":158,"workload":10,"cost_metric":159},"H200 SXM",11,141,3.5,2555,[150],"lambda",4800,989,40,0.5907,"balanced","tok\u002Fs","110.4 tok\u002Fs",110.4,{"effective_value":160,"effective_unit":156,"capacity_factor":161},77.3,0.7,{"gpu_model":163,"gpu_model_id":164,"gpu_count":34,"total_vram_gb":165,"hour_price":166,"monthly_cost":167,"providers":168,"bandwidth_gbps":41,"fp16_tflops":41,"interconnect_score":153,"parallel_efficiency":170,"score":171,"tier":172,"perf_metric":156,"perf_label":173,"perf_value":41,"workload":10,"cost_metric":174},"Tesla V100",17,64,0.0272,79.42,[169],"vast",0.6,0.545,"cheapest","0 tok\u002Fs",{"effective_value":41,"effective_unit":156,"capacity_factor":161},{"gpu_model":144,"gpu_model_id":145,"gpu_count":44,"total_vram_gb":176,"hour_price":147,"monthly_cost":177,"providers":178,"bandwidth_gbps":151,"fp16_tflops":152,"interconnect_score":153,"parallel_efficiency":179,"score":180,"tier":155,"perf_metric":156,"perf_label":181,"perf_value":182,"workload":10,"cost_metric":183},1128,20440,[150],0.4,0.2197,"353.3 tok\u002Fs",353.3,{"effective_value":184,"effective_unit":156,"capacity_factor":161},247.3,{"gpu_model":186,"gpu_model_id":187,"gpu_count":40,"total_vram_gb":188,"hour_price":189,"monthly_cost":190,"providers":191,"bandwidth_gbps":194,"fp16_tflops":195,"interconnect_score":153,"parallel_efficiency":40,"score":196,"tier":155,"perf_metric":156,"perf_label":197,"perf_value":198,"workload":10,"cost_metric":199},"A100 80GB",9,80,1.2,876,[169,192,193,150],"tensordock","runpod",2039,77.97,0.6755,"46.9 tok\u002Fs",46.9,{"effective_value":200,"effective_unit":156,"capacity_factor":161},32.8,[],false,20,"value",[206,210,218,222,230],{"name":186,"vram_gb":188,"cheapest_hour":189,"cost":207,"provider":169,"estimated_tps":198},{"hourly":189,"daily":208,"monthly":190,"yearly":209},28.8,10512,{"name":211,"vram_gb":188,"cheapest_hour":212,"cost":213,"provider":169,"estimated_tps":217},"H100 SXM",1.6022,{"hourly":212,"daily":214,"monthly":215,"yearly":216},38.45,1169.61,14035.27,77,{"name":144,"vram_gb":146,"cheapest_hour":147,"cost":219,"provider":150,"estimated_tps":158},{"hourly":147,"daily":220,"monthly":148,"yearly":221},84,30660,{"name":223,"vram_gb":224,"cheapest_hour":225,"cost":226,"provider":169,"estimated_tps":41},"B200",180,4.1271,{"hourly":225,"daily":227,"monthly":228,"yearly":229},99.05,3012.78,36153.4,{"name":231,"vram_gb":224,"cheapest_hour":10,"cost":10,"provider":10,"estimated_tps":232},"B200 SXM",179.4,{"posts_count":41,"benchmarks_count":41}]