[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-devstral-2-123b":3},{"id":4,"slug":5,"name":6,"hf_id":7,"organization":8,"architecture":9,"architecture_config":8,"parameters":10,"parameters_label":11,"context_length":12,"license":13,"downloads":14,"likes":14,"status":15,"description":8,"huggingface_url":16,"created_at":17,"seo":18,"category":21,"family":8,"score":14,"variants":24,"quantizations":25,"requirements":51,"gpu_recommendations":136,"fitting_gpus":213,"community":229},118,"devstral-2-123b","Devstral 2 123B","mistralai\u002FDevstral-2-123B",null,"transformer",123000000000,"123B",131072,"MIT",0,"published","https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FDevstral-2-123B","2026-08-12T03:11:12+00:00",{"title":6,"description":19,"image":8,"robots":20,"canonical_url":8},"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":22,"slug":23},"Coding AI","coding",[],[26,31,35,40,44,48],{"format":27,"bits":28,"size_factor":29,"quality_loss":30},"AWQ",4,0.263,0.03,{"format":32,"bits":33,"size_factor":34,"quality_loss":14},"FP16",16,1,{"format":36,"bits":37,"size_factor":38,"quality_loss":39},"FP8",8,0.5,0.01,{"format":41,"bits":28,"size_factor":42,"quality_loss":43},"GGUF",0.25,0.05,{"format":45,"bits":28,"size_factor":46,"quality_loss":47},"GPTQ",0.275,0.04,{"format":49,"bits":37,"size_factor":38,"quality_loss":50},"INT8",0.02,{"FP16":52,"FP8":76,"INT8":90,"AWQ":94,"GPTQ":108,"GGUF":122},{"minimum":53,"production":62,"recommended":69},{"scenario":54,"context_length":55,"batch_size":34,"weight_gb":56,"kv_cache_gb":38,"vram_gb":57,"ram_gb":58,"disk_gb":59,"requirement_source":60,"confidence":61},"minimum",4096,229.11,294.77,442.15,396.4,"estimated",0.95,{"scenario":63,"context_length":64,"batch_size":28,"weight_gb":56,"kv_cache_gb":65,"vram_gb":66,"ram_gb":67,"disk_gb":68,"requirement_source":60,"confidence":61},"production",16384,64,433.77,650.65,461.4,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":56,"kv_cache_gb":37,"vram_gb":73,"ram_gb":74,"disk_gb":75,"requirement_source":60,"confidence":61},"recommended",8192,2,323.69,485.54,422.4,{"minimum":77,"production":82,"recommended":86},{"scenario":54,"context_length":55,"batch_size":34,"weight_gb":78,"kv_cache_gb":38,"vram_gb":79,"ram_gb":80,"disk_gb":81,"requirement_source":60,"confidence":61},114.55,149.86,224.79,217.7,{"scenario":63,"context_length":64,"batch_size":28,"weight_gb":78,"kv_cache_gb":65,"vram_gb":83,"ram_gb":84,"disk_gb":85,"requirement_source":60,"confidence":61},275.68,413.52,282.7,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":78,"kv_cache_gb":37,"vram_gb":87,"ram_gb":88,"disk_gb":89,"requirement_source":60,"confidence":61},172.2,258.29,243.7,{"minimum":91,"production":92,"recommended":93},{"scenario":54,"context_length":55,"batch_size":34,"weight_gb":78,"kv_cache_gb":38,"vram_gb":79,"ram_gb":80,"disk_gb":81,"requirement_source":60,"confidence":61},{"scenario":63,"context_length":64,"batch_size":28,"weight_gb":78,"kv_cache_gb":65,"vram_gb":83,"ram_gb":84,"disk_gb":85,"requirement_source":60,"confidence":61},{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":78,"kv_cache_gb":37,"vram_gb":87,"ram_gb":88,"disk_gb":89,"requirement_source":60,"confidence":61},{"minimum":95,"production":100,"recommended":104},{"scenario":54,"context_length":55,"batch_size":34,"weight_gb":96,"kv_cache_gb":38,"vram_gb":97,"ram_gb":98,"disk_gb":99,"requirement_source":60,"confidence":61},59.42,80.12,120.18,131.7,{"scenario":63,"context_length":64,"batch_size":28,"weight_gb":96,"kv_cache_gb":65,"vram_gb":101,"ram_gb":102,"disk_gb":103,"requirement_source":60,"confidence":61},199.61,299.41,196.7,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":96,"kv_cache_gb":37,"vram_gb":105,"ram_gb":106,"disk_gb":107,"requirement_source":60,"confidence":61},99.29,148.93,157.7,{"minimum":109,"production":114,"recommended":118},{"scenario":54,"context_length":55,"batch_size":34,"weight_gb":110,"kv_cache_gb":38,"vram_gb":111,"ram_gb":112,"disk_gb":113,"requirement_source":60,"confidence":61},60.14,81.03,121.54,132.82,{"scenario":63,"context_length":64,"batch_size":28,"weight_gb":110,"kv_cache_gb":65,"vram_gb":115,"ram_gb":116,"disk_gb":117,"requirement_source":60,"confidence":61},200.59,300.89,197.82,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":110,"kv_cache_gb":37,"vram_gb":119,"ram_gb":120,"disk_gb":121,"requirement_source":60,"confidence":61},100.24,150.35,158.82,{"minimum":123,"production":128,"recommended":132},{"scenario":54,"context_length":55,"batch_size":34,"weight_gb":124,"kv_cache_gb":38,"vram_gb":125,"ram_gb":126,"disk_gb":127,"requirement_source":60,"confidence":61},61.57,82.84,124.26,135.05,{"scenario":63,"context_length":64,"batch_size":28,"weight_gb":124,"kv_cache_gb":65,"vram_gb":129,"ram_gb":130,"disk_gb":131,"requirement_source":60,"confidence":61},202.57,303.85,200.05,{"scenario":70,"context_length":71,"batch_size":72,"weight_gb":124,"kv_cache_gb":37,"vram_gb":133,"ram_gb":134,"disk_gb":135,"requirement_source":60,"confidence":61},102.13,153.19,161.05,{"recommended":137,"cheapest":157,"performance":170,"min_complexity":179,"cluster_options":188,"no_match":210,"required_vram_gb":105,"candidate_count":211,"usage":212},{"gpu_model":138,"gpu_model_id":139,"gpu_count":28,"total_vram_gb":140,"hour_price":141,"monthly_cost":142,"providers":143,"bandwidth_gbps":145,"fp16_tflops":146,"interconnect_score":147,"parallel_efficiency":148,"score":149,"tier":150,"perf_metric":151,"perf_label":152,"perf_value":153,"workload":8,"cost_metric":154},"H200 SXM",11,564,3.5,10220,[144],"lambda",4800,989,40,0.6,0.3163,"balanced","tok\u002Fs","193.9 tok\u002Fs",193.9,{"effective_value":155,"effective_unit":151,"capacity_factor":156},135.7,0.7,{"gpu_model":158,"gpu_model_id":159,"gpu_count":37,"total_vram_gb":160,"hour_price":161,"monthly_cost":162,"providers":163,"bandwidth_gbps":14,"fp16_tflops":14,"interconnect_score":147,"parallel_efficiency":165,"score":166,"tier":167,"perf_metric":151,"perf_label":168,"perf_value":14,"workload":8,"cost_metric":169},"Tesla V100",17,128,0.0272,158.85,[164],"vast",0.4,0.5498,"cheapest","0 tok\u002Fs",{"effective_value":14,"effective_unit":151,"capacity_factor":156},{"gpu_model":138,"gpu_model_id":139,"gpu_count":37,"total_vram_gb":171,"hour_price":141,"monthly_cost":172,"providers":173,"bandwidth_gbps":145,"fp16_tflops":146,"interconnect_score":147,"parallel_efficiency":165,"score":174,"tier":150,"perf_metric":151,"perf_label":175,"perf_value":176,"workload":8,"cost_metric":177},1128,20440,[144],0.2197,"258.5 tok\u002Fs",258.5,{"effective_value":178,"effective_unit":151,"capacity_factor":156},181,{"gpu_model":138,"gpu_model_id":139,"gpu_count":34,"total_vram_gb":180,"hour_price":141,"monthly_cost":181,"providers":182,"bandwidth_gbps":145,"fp16_tflops":146,"interconnect_score":147,"parallel_efficiency":34,"score":183,"tier":150,"perf_metric":151,"perf_label":184,"perf_value":185,"workload":8,"cost_metric":186},141,2555,[144],0.6588,"80.8 tok\u002Fs",80.8,{"effective_value":187,"effective_unit":151,"capacity_factor":156},56.6,[189,201],{"gpu_model":190,"gpu_model_id":191,"gpu_count":37,"total_vram_gb":192,"hour_price":193,"monthly_cost":194,"providers":195,"bandwidth_gbps":196,"fp16_tflops":197,"interconnect_score":147,"parallel_efficiency":165,"insufficient":198,"vram_shortfall_gb":199,"score":200,"tier":167},"RTX 5070",28,96,0.0101,58.98,[164],672,61,true,3.29,0.5208,{"gpu_model":202,"gpu_model_id":203,"gpu_count":37,"total_vram_gb":65,"hour_price":204,"monthly_cost":205,"providers":206,"bandwidth_gbps":207,"fp16_tflops":203,"interconnect_score":147,"parallel_efficiency":165,"insufficient":198,"vram_shortfall_gb":208,"score":209,"tier":150},"RTX 5060 Ti",24,0.0544,317.7,[164],448,35.29,0.5103,false,14,"value",[214,218,226],{"name":138,"vram_gb":180,"cheapest_hour":141,"cost":215,"provider":144,"estimated_tps":185},{"hourly":141,"daily":216,"monthly":181,"yearly":217},84,30660,{"name":219,"vram_gb":220,"cheapest_hour":221,"cost":222,"provider":164,"estimated_tps":14},"B200",180,4.1271,{"hourly":221,"daily":223,"monthly":224,"yearly":225},99.05,3012.78,36153.4,{"name":227,"vram_gb":220,"cheapest_hour":8,"cost":8,"provider":8,"estimated_tps":228},"B200 SXM",131.3,{"posts_count":14,"benchmarks_count":14}]