[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-leanstral":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":139,"fitting_gpus":214,"community":230},199,"leanstral","Leanstral 1.5","mistralai\u002FLeanstral-1.5-119B-A6B","mistralai","moe_transformer",null,119000000000,"119B",262144,"MIT",291,212,"published","https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FLeanstral-1.5-119B-A6B","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",0.0023,[],[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":79,"INT8":93,"AWQ":97,"GPTQ":111,"GGUF":125},{"minimum":57,"production":66,"recommended":73},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":60,"kv_cache_gb":46,"vram_gb":61,"ram_gb":62,"disk_gb":63,"requirement_source":64,"confidence":65},"minimum",2048,221.65,283.97,425.95,384.78,"estimated",0.95,{"scenario":67,"context_length":68,"batch_size":69,"weight_gb":60,"kv_cache_gb":41,"vram_gb":70,"ram_gb":71,"disk_gb":72,"requirement_source":64,"confidence":65},"production",8192,2,327.48,491.23,449.78,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":60,"kv_cache_gb":37,"vram_gb":76,"ram_gb":77,"disk_gb":78,"requirement_source":64,"confidence":65},"recommended",4096,298.89,448.33,410.78,{"minimum":80,"production":85,"recommended":89},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":81,"kv_cache_gb":46,"vram_gb":82,"ram_gb":83,"disk_gb":84,"requirement_source":64,"confidence":65},110.83,143.77,215.66,211.89,{"scenario":67,"context_length":68,"batch_size":69,"weight_gb":81,"kv_cache_gb":41,"vram_gb":86,"ram_gb":87,"disk_gb":88,"requirement_source":64,"confidence":65},174.54,261.81,276.89,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":81,"kv_cache_gb":37,"vram_gb":90,"ram_gb":91,"disk_gb":92,"requirement_source":64,"confidence":65},152.32,228.48,237.89,{"minimum":94,"production":95,"recommended":96},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":81,"kv_cache_gb":46,"vram_gb":82,"ram_gb":83,"disk_gb":84,"requirement_source":64,"confidence":65},{"scenario":67,"context_length":68,"batch_size":69,"weight_gb":81,"kv_cache_gb":41,"vram_gb":86,"ram_gb":87,"disk_gb":88,"requirement_source":64,"confidence":65},{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":81,"kv_cache_gb":37,"vram_gb":90,"ram_gb":91,"disk_gb":92,"requirement_source":64,"confidence":65},{"minimum":98,"production":103,"recommended":107},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":99,"kv_cache_gb":46,"vram_gb":100,"ram_gb":101,"disk_gb":102,"requirement_source":64,"confidence":65},57.49,76.3,114.45,128.69,{"scenario":67,"context_length":68,"batch_size":69,"weight_gb":99,"kv_cache_gb":41,"vram_gb":104,"ram_gb":105,"disk_gb":106,"requirement_source":64,"confidence":65},100.94,151.41,193.69,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":99,"kv_cache_gb":37,"vram_gb":108,"ram_gb":109,"disk_gb":110,"requirement_source":64,"confidence":65},81.78,122.67,154.69,{"minimum":112,"production":117,"recommended":121},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":113,"kv_cache_gb":46,"vram_gb":114,"ram_gb":115,"disk_gb":116,"requirement_source":64,"confidence":65},58.18,77.18,115.77,129.77,{"scenario":67,"context_length":68,"batch_size":69,"weight_gb":113,"kv_cache_gb":41,"vram_gb":118,"ram_gb":119,"disk_gb":120,"requirement_source":64,"confidence":65},101.89,152.84,194.77,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":113,"kv_cache_gb":37,"vram_gb":122,"ram_gb":123,"disk_gb":124,"requirement_source":64,"confidence":65},82.7,124.05,155.77,{"minimum":126,"production":131,"recommended":135},{"scenario":58,"context_length":59,"batch_size":37,"weight_gb":127,"kv_cache_gb":46,"vram_gb":128,"ram_gb":129,"disk_gb":130,"requirement_source":64,"confidence":65},59.57,78.93,118.4,131.93,{"scenario":67,"context_length":68,"batch_size":69,"weight_gb":127,"kv_cache_gb":41,"vram_gb":132,"ram_gb":133,"disk_gb":134,"requirement_source":64,"confidence":65},103.81,155.71,196.93,{"scenario":74,"context_length":75,"batch_size":37,"weight_gb":127,"kv_cache_gb":37,"vram_gb":136,"ram_gb":137,"disk_gb":138,"requirement_source":64,"confidence":65},84.53,126.8,157.93,{"recommended":140,"cheapest":159,"performance":176,"min_complexity":185,"cluster_options":199,"no_match":211,"required_vram_gb":108,"candidate_count":212,"usage":213},{"gpu_model":141,"gpu_model_id":142,"gpu_count":37,"total_vram_gb":143,"hour_price":144,"monthly_cost":145,"providers":146,"bandwidth_gbps":148,"fp16_tflops":149,"interconnect_score":150,"parallel_efficiency":37,"score":151,"tier":152,"perf_metric":153,"perf_label":154,"perf_value":155,"workload":10,"cost_metric":156},"H200 SXM",11,141,3.5,2555,[147],"lambda",4800,989,40,0.6257,"balanced","tok\u002Fs","83.5 tok\u002Fs",83.5,{"effective_value":157,"effective_unit":153,"capacity_factor":158},58.4,0.7,{"gpu_model":160,"gpu_model_id":161,"gpu_count":41,"total_vram_gb":162,"hour_price":163,"monthly_cost":164,"providers":165,"bandwidth_gbps":167,"fp16_tflops":168,"interconnect_score":150,"parallel_efficiency":169,"score":170,"tier":171,"perf_metric":153,"perf_label":172,"perf_value":173,"workload":10,"cost_metric":174},"RTX 5070",28,96,0.0101,58.98,[166],"vast",672,61,0.4,0.547,"cheapest","37.4 tok\u002Fs",37.4,{"effective_value":175,"effective_unit":153,"capacity_factor":158},26.2,{"gpu_model":141,"gpu_model_id":142,"gpu_count":41,"total_vram_gb":177,"hour_price":144,"monthly_cost":178,"providers":179,"bandwidth_gbps":148,"fp16_tflops":149,"interconnect_score":150,"parallel_efficiency":169,"score":180,"tier":152,"perf_metric":153,"perf_label":181,"perf_value":182,"workload":10,"cost_metric":183},1128,20440,[147],0.2197,"267.2 tok\u002Fs",267.2,{"effective_value":184,"effective_unit":153,"capacity_factor":158},187,{"gpu_model":186,"gpu_model_id":187,"gpu_count":69,"total_vram_gb":162,"hour_price":188,"monthly_cost":189,"providers":190,"bandwidth_gbps":191,"fp16_tflops":192,"interconnect_score":150,"parallel_efficiency":193,"score":194,"tier":152,"perf_metric":153,"perf_label":195,"perf_value":196,"workload":10,"cost_metric":197},"RTX A6000",6,0.2874,419.6,[166,147],768,77,0.8,0.6204,"21.4 tok\u002Fs",21.4,{"effective_value":198,"effective_unit":153,"capacity_factor":158},15,[200],{"gpu_model":201,"gpu_model_id":202,"gpu_count":41,"total_vram_gb":203,"hour_price":204,"monthly_cost":205,"providers":206,"bandwidth_gbps":207,"fp16_tflops":202,"interconnect_score":150,"parallel_efficiency":169,"insufficient":208,"vram_shortfall_gb":209,"score":210,"tier":152},"RTX 5060 Ti",24,64,0.0544,317.7,[166],448,true,17.78,0.5103,false,17,"value",[215,219,227],{"name":141,"vram_gb":143,"cheapest_hour":144,"cost":216,"provider":147,"estimated_tps":155},{"hourly":144,"daily":217,"monthly":145,"yearly":218},84,30660,{"name":220,"vram_gb":221,"cheapest_hour":222,"cost":223,"provider":166,"estimated_tps":38},"B200",180,4.1271,{"hourly":222,"daily":224,"monthly":225,"yearly":226},99.05,3012.78,36153.4,{"name":228,"vram_gb":221,"cheapest_hour":10,"cost":10,"provider":10,"estimated_tps":229},"B200 SXM",135.7,{"posts_count":38,"benchmarks_count":38}]