[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-granite-code-20b":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":10,"score":26,"variants":27,"quantizations":28,"requirements":55,"gpu_recommendations":138,"fitting_gpus":205,"community":240},175,"granite-code-20b","Granite Code 20B","ibm-granite\u002Fgranite-20b-code-instruct","ibm-granite","transformer",null,20000000000,"20B",131072,"MIT",9891,44,"published","https:\u002F\u002Fhuggingface.co\u002Fibm-granite\u002Fgranite-20b-code-instruct","2026-08-12T03:11:15+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},"Coding AI","coding",0.0005,[],[29,34,39,44,48,52],{"format":30,"bits":31,"size_factor":32,"quality_loss":33},"AWQ",4,0.263,0.03,{"format":35,"bits":36,"size_factor":37,"quality_loss":38},"FP16",16,1,0,{"format":40,"bits":41,"size_factor":42,"quality_loss":43},"FP8",8,0.5,0.01,{"format":45,"bits":31,"size_factor":46,"quality_loss":47},"GGUF",0.25,0.05,{"format":49,"bits":31,"size_factor":50,"quality_loss":51},"GPTQ",0.275,0.04,{"format":53,"bits":41,"size_factor":42,"quality_loss":54},"INT8",0.02,{"FP16":56,"FP8":80,"INT8":94,"AWQ":98,"GPTQ":112,"GGUF":125},{"minimum":57,"production":66,"recommended":73},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":60,"kv_cache_gb":42,"vram_gb":61,"ram_gb":62,"disk_gb":63,"requirement_source":64,"confidence":65},"minimum",4096,37.25,52.07,78.11,97.11,"estimated",0.95,{"scenario":67,"context_length":68,"batch_size":31,"weight_gb":60,"kv_cache_gb":69,"vram_gb":70,"ram_gb":71,"disk_gb":72,"requirement_source":64,"confidence":65},"production",16384,64,169.01,253.51,162.11,{"scenario":74,"context_length":75,"batch_size":76,"weight_gb":60,"kv_cache_gb":41,"vram_gb":77,"ram_gb":78,"disk_gb":79,"requirement_source":64,"confidence":65},"recommended",8192,2,69.97,104.95,123.11,{"minimum":81,"production":86,"recommended":90},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":82,"kv_cache_gb":42,"vram_gb":83,"ram_gb":84,"disk_gb":85,"requirement_source":64,"confidence":65},18.63,28.51,42.77,68.06,{"scenario":67,"context_length":68,"batch_size":31,"weight_gb":82,"kv_cache_gb":69,"vram_gb":87,"ram_gb":88,"disk_gb":89,"requirement_source":64,"confidence":65},143.3,214.96,133.06,{"scenario":74,"context_length":75,"batch_size":76,"weight_gb":82,"kv_cache_gb":41,"vram_gb":91,"ram_gb":92,"disk_gb":93,"requirement_source":64,"confidence":65},45.33,68,94.06,{"minimum":95,"production":96,"recommended":97},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":82,"kv_cache_gb":42,"vram_gb":83,"ram_gb":84,"disk_gb":85,"requirement_source":64,"confidence":65},{"scenario":67,"context_length":68,"batch_size":31,"weight_gb":82,"kv_cache_gb":69,"vram_gb":87,"ram_gb":88,"disk_gb":89,"requirement_source":64,"confidence":65},{"scenario":74,"context_length":75,"batch_size":76,"weight_gb":82,"kv_cache_gb":41,"vram_gb":91,"ram_gb":92,"disk_gb":93,"requirement_source":64,"confidence":65},{"minimum":99,"production":104,"recommended":108},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":100,"kv_cache_gb":42,"vram_gb":101,"ram_gb":102,"disk_gb":103,"requirement_source":64,"confidence":65},9.66,17.17,32,54.07,{"scenario":67,"context_length":68,"batch_size":31,"weight_gb":100,"kv_cache_gb":69,"vram_gb":105,"ram_gb":106,"disk_gb":107,"requirement_source":64,"confidence":65},130.93,196.4,119.07,{"scenario":74,"context_length":75,"batch_size":76,"weight_gb":100,"kv_cache_gb":41,"vram_gb":109,"ram_gb":110,"disk_gb":111,"requirement_source":64,"confidence":65},33.48,50.22,80.07,{"minimum":113,"production":117,"recommended":121},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":114,"kv_cache_gb":42,"vram_gb":115,"ram_gb":102,"disk_gb":116,"requirement_source":64,"confidence":65},9.78,17.32,54.26,{"scenario":67,"context_length":68,"batch_size":31,"weight_gb":114,"kv_cache_gb":69,"vram_gb":118,"ram_gb":119,"disk_gb":120,"requirement_source":64,"confidence":65},131.09,196.64,119.26,{"scenario":74,"context_length":75,"batch_size":76,"weight_gb":114,"kv_cache_gb":41,"vram_gb":122,"ram_gb":123,"disk_gb":124,"requirement_source":64,"confidence":65},33.63,50.45,80.26,{"minimum":126,"production":130,"recommended":134},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":127,"kv_cache_gb":42,"vram_gb":128,"ram_gb":102,"disk_gb":129,"requirement_source":64,"confidence":65},10.01,17.61,54.62,{"scenario":67,"context_length":68,"batch_size":31,"weight_gb":127,"kv_cache_gb":69,"vram_gb":131,"ram_gb":132,"disk_gb":133,"requirement_source":64,"confidence":65},131.42,197.12,119.62,{"scenario":74,"context_length":75,"batch_size":76,"weight_gb":127,"kv_cache_gb":41,"vram_gb":135,"ram_gb":136,"disk_gb":137,"requirement_source":64,"confidence":65},33.94,50.91,80.62,{"recommended":139,"cheapest":161,"performance":172,"min_complexity":187,"cluster_options":201,"no_match":202,"required_vram_gb":109,"candidate_count":203,"usage":204},{"gpu_model":140,"gpu_model_id":141,"gpu_count":37,"total_vram_gb":142,"hour_price":143,"monthly_cost":144,"providers":145,"bandwidth_gbps":150,"fp16_tflops":151,"interconnect_score":152,"parallel_efficiency":37,"score":153,"tier":154,"perf_metric":155,"perf_label":156,"perf_value":157,"workload":10,"cost_metric":158},"H100 SXM",10,80,1.6022,1169.61,[146,147,148,149],"vast","tensordock","lambda","runpod",3350,66.9,40,0.6012,"balanced","tok\u002Fs","346.8 tok\u002Fs",346.8,{"effective_value":159,"effective_unit":155,"capacity_factor":160},242.8,0.7,{"gpu_model":162,"gpu_model_id":163,"gpu_count":31,"total_vram_gb":69,"hour_price":164,"monthly_cost":165,"providers":166,"bandwidth_gbps":38,"fp16_tflops":38,"interconnect_score":152,"parallel_efficiency":167,"score":168,"tier":169,"perf_metric":155,"perf_label":170,"perf_value":38,"workload":10,"cost_metric":171},"Tesla V100",17,0.0272,79.42,[146],0.6,0.5378,"cheapest","0 tok\u002Fs",{"effective_value":38,"effective_unit":155,"capacity_factor":160},{"gpu_model":173,"gpu_model_id":174,"gpu_count":41,"total_vram_gb":175,"hour_price":176,"monthly_cost":177,"providers":178,"bandwidth_gbps":179,"fp16_tflops":180,"interconnect_score":152,"parallel_efficiency":181,"score":182,"tier":154,"perf_metric":155,"perf_label":183,"perf_value":184,"workload":10,"cost_metric":185},"H200 SXM",11,1128,3.5,20440,[148],4800,989,0.4,0.2197,"1590.1 tok\u002Fs",1590.1,{"effective_value":186,"effective_unit":155,"capacity_factor":160},1113.1,{"gpu_model":188,"gpu_model_id":189,"gpu_count":37,"total_vram_gb":190,"hour_price":191,"monthly_cost":192,"providers":193,"bandwidth_gbps":194,"fp16_tflops":195,"interconnect_score":152,"parallel_efficiency":37,"score":196,"tier":169,"perf_metric":155,"perf_label":197,"perf_value":198,"workload":10,"cost_metric":199},"RTX A6000",6,48,0.3001,219.07,[146,148],768,77,0.6769,"79.5 tok\u002Fs",79.5,{"effective_value":200,"effective_unit":155,"capacity_factor":160},55.7,[],false,23,"value",[206,210,218,225,232],{"name":188,"vram_gb":190,"cheapest_hour":191,"cost":207,"provider":146,"estimated_tps":198},{"hourly":191,"daily":208,"monthly":192,"yearly":209},7.2,2628.88,{"name":211,"vram_gb":190,"cheapest_hour":212,"cost":213,"provider":146,"estimated_tps":217},"L40S",0.4669,{"hourly":212,"daily":214,"monthly":215,"yearly":216},11.21,340.84,4090.04,89.4,{"name":219,"vram_gb":190,"cheapest_hour":220,"cost":221,"provider":146,"estimated_tps":38},"RTX PRO 6000 WS",0.8022,{"hourly":220,"daily":222,"monthly":223,"yearly":224},19.25,585.61,7027.27,{"name":226,"vram_gb":190,"cheapest_hour":227,"cost":228,"provider":146,"estimated_tps":38},"RTX PRO 6000 S",1.0131,{"hourly":227,"daily":229,"monthly":230,"yearly":231},24.31,739.56,8874.76,{"name":233,"vram_gb":142,"cheapest_hour":234,"cost":235,"provider":146,"estimated_tps":239},"A100 80GB",1.2,{"hourly":234,"daily":236,"monthly":237,"yearly":238},28.8,876,10512,211.1,{"posts_count":38,"benchmarks_count":38}]