[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-prompt-guard-86m":3},{"id":4,"slug":5,"name":6,"hf_id":6,"organization":7,"architecture":8,"architecture_config":8,"parameters":9,"parameters_label":10,"context_length":8,"license":8,"downloads":11,"likes":12,"status":13,"description":8,"huggingface_url":14,"created_at":15,"seo":16,"category":19,"family":22,"score":25,"variants":26,"quantizations":27,"requirements":54,"gpu_recommendations":117,"fitting_gpus":182,"community":218},502,"prompt-guard-86m","meta-llama\u002FPrompt-Guard-86M","meta-llama",null,86000000,"86M",4444646,390,"published","https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FPrompt-Guard-86M","2026-08-13T18:12:25+00:00",{"title":6,"description":17,"image":8,"robots":18,"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":20,"slug":21},"LLM","llm",{"name":23,"slug":24},"Llama","llama",0.065,[],[28,33,38,43,47,51],{"format":29,"bits":30,"size_factor":31,"quality_loss":32},"AWQ",4,0.263,0.03,{"format":34,"bits":35,"size_factor":36,"quality_loss":37},"FP16",16,1,0,{"format":39,"bits":40,"size_factor":41,"quality_loss":42},"FP8",8,0.5,0.01,{"format":44,"bits":30,"size_factor":45,"quality_loss":46},"GGUF",0.25,0.05,{"format":48,"bits":30,"size_factor":49,"quality_loss":50},"GPTQ",0.275,0.04,{"format":52,"bits":40,"size_factor":41,"quality_loss":53},"INT8",0.02,{"FP16":55,"FP8":78,"INT8":90,"AWQ":94,"GPTQ":105,"GGUF":113},{"minimum":56,"production":65,"recommended":72},{"scenario":57,"context_length":58,"batch_size":36,"weight_gb":59,"kv_cache_gb":41,"vram_gb":60,"ram_gb":61,"disk_gb":62,"requirement_source":63,"confidence":64},"minimum",4096,0.16,5.15,32,39.25,"inferred",0.95,{"scenario":66,"context_length":67,"batch_size":30,"weight_gb":59,"kv_cache_gb":68,"vram_gb":69,"ram_gb":70,"disk_gb":71,"requirement_source":63,"confidence":64},"production",16384,64,117.82,176.73,104.25,{"scenario":73,"context_length":74,"batch_size":75,"weight_gb":59,"kv_cache_gb":40,"vram_gb":76,"ram_gb":61,"disk_gb":77,"requirement_source":63,"confidence":64},"recommended",8192,2,20.91,65.25,{"minimum":79,"production":83,"recommended":87},{"scenario":57,"context_length":58,"batch_size":36,"weight_gb":80,"kv_cache_gb":41,"vram_gb":81,"ram_gb":61,"disk_gb":82,"requirement_source":63,"confidence":64},0.08,5.05,39.12,{"scenario":66,"context_length":67,"batch_size":30,"weight_gb":80,"kv_cache_gb":68,"vram_gb":84,"ram_gb":85,"disk_gb":86,"requirement_source":63,"confidence":64},117.71,176.57,104.12,{"scenario":73,"context_length":74,"batch_size":75,"weight_gb":80,"kv_cache_gb":40,"vram_gb":88,"ram_gb":61,"disk_gb":89,"requirement_source":63,"confidence":64},20.81,65.12,{"minimum":91,"production":92,"recommended":93},{"scenario":57,"context_length":58,"batch_size":36,"weight_gb":80,"kv_cache_gb":41,"vram_gb":81,"ram_gb":61,"disk_gb":82,"requirement_source":63,"confidence":64},{"scenario":66,"context_length":67,"batch_size":30,"weight_gb":80,"kv_cache_gb":68,"vram_gb":84,"ram_gb":85,"disk_gb":86,"requirement_source":63,"confidence":64},{"scenario":73,"context_length":74,"batch_size":75,"weight_gb":80,"kv_cache_gb":40,"vram_gb":88,"ram_gb":61,"disk_gb":89,"requirement_source":63,"confidence":64},{"minimum":95,"production":98,"recommended":102},{"scenario":57,"context_length":58,"batch_size":36,"weight_gb":50,"kv_cache_gb":41,"vram_gb":96,"ram_gb":61,"disk_gb":97,"requirement_source":63,"confidence":64},5,39.06,{"scenario":66,"context_length":67,"batch_size":30,"weight_gb":50,"kv_cache_gb":68,"vram_gb":99,"ram_gb":100,"disk_gb":101,"requirement_source":63,"confidence":64},117.66,176.49,104.06,{"scenario":73,"context_length":74,"batch_size":75,"weight_gb":50,"kv_cache_gb":40,"vram_gb":103,"ram_gb":61,"disk_gb":104,"requirement_source":63,"confidence":64},20.75,65.06,{"minimum":106,"production":108,"recommended":110},{"scenario":57,"context_length":58,"batch_size":36,"weight_gb":50,"kv_cache_gb":41,"vram_gb":96,"ram_gb":61,"disk_gb":107,"requirement_source":63,"confidence":64},39.07,{"scenario":66,"context_length":67,"batch_size":30,"weight_gb":50,"kv_cache_gb":68,"vram_gb":99,"ram_gb":100,"disk_gb":109,"requirement_source":63,"confidence":64},104.07,{"scenario":73,"context_length":74,"batch_size":75,"weight_gb":50,"kv_cache_gb":40,"vram_gb":111,"ram_gb":61,"disk_gb":112,"requirement_source":63,"confidence":64},20.76,65.07,{"minimum":114,"production":115,"recommended":116},{"scenario":57,"context_length":58,"batch_size":36,"weight_gb":50,"kv_cache_gb":41,"vram_gb":96,"ram_gb":61,"disk_gb":107,"requirement_source":63,"confidence":64},{"scenario":66,"context_length":67,"batch_size":30,"weight_gb":50,"kv_cache_gb":68,"vram_gb":99,"ram_gb":100,"disk_gb":109,"requirement_source":63,"confidence":64},{"scenario":73,"context_length":74,"batch_size":75,"weight_gb":50,"kv_cache_gb":40,"vram_gb":111,"ram_gb":61,"disk_gb":112,"requirement_source":63,"confidence":64},{"recommended":118,"cheapest":136,"performance":146,"min_complexity":163,"cluster_options":178,"no_match":179,"required_vram_gb":103,"candidate_count":180,"usage":181},{"gpu_model":119,"gpu_model_id":75,"gpu_count":36,"total_vram_gb":61,"hour_price":120,"monthly_cost":121,"providers":122,"bandwidth_gbps":125,"fp16_tflops":126,"interconnect_score":127,"parallel_efficiency":36,"score":128,"tier":129,"perf_metric":130,"perf_label":131,"perf_value":132,"workload":8,"cost_metric":133},"RTX 5090",0.3222,235.21,[123,124],"vast","runpod",1792,209,40,0.6708,"cheapest","tok\u002Fs","44800 tok\u002Fs",44800,{"effective_value":134,"effective_unit":130,"capacity_factor":135},31360,0.7,{"gpu_model":137,"gpu_model_id":138,"gpu_count":75,"total_vram_gb":61,"hour_price":139,"monthly_cost":140,"providers":141,"bandwidth_gbps":37,"fp16_tflops":37,"interconnect_score":127,"parallel_efficiency":142,"score":143,"tier":129,"perf_metric":130,"perf_label":144,"perf_value":37,"workload":8,"cost_metric":145},"Tesla V100",17,0.0272,39.71,[123],0.8,0.6196,"0 tok\u002Fs",{"effective_value":37,"effective_unit":130,"capacity_factor":135},{"gpu_model":147,"gpu_model_id":148,"gpu_count":40,"total_vram_gb":149,"hour_price":150,"monthly_cost":151,"providers":152,"bandwidth_gbps":154,"fp16_tflops":155,"interconnect_score":127,"parallel_efficiency":156,"score":157,"tier":158,"perf_metric":130,"perf_label":159,"perf_value":160,"workload":8,"cost_metric":161},"H200 SXM",11,1128,3.5,20440,[153],"lambda",4800,989,0.4,0.2197,"balanced","384000 tok\u002Fs",384000,{"effective_value":162,"effective_unit":130,"capacity_factor":135},268800,{"gpu_model":164,"gpu_model_id":165,"gpu_count":36,"total_vram_gb":166,"hour_price":167,"monthly_cost":168,"providers":169,"bandwidth_gbps":171,"fp16_tflops":172,"interconnect_score":127,"parallel_efficiency":36,"score":173,"tier":129,"perf_metric":130,"perf_label":174,"perf_value":175,"workload":8,"cost_metric":176},"RTX 3090",3,24,0.0678,49.49,[123,170],"tensordock",936,71,0.6651,"23400 tok\u002Fs",23400,{"effective_value":177,"effective_unit":130,"capacity_factor":135},16380,[],false,25,"value",[183,187,195,202,210],{"name":164,"vram_gb":166,"cheapest_hour":167,"cost":184,"provider":123,"estimated_tps":175},{"hourly":167,"daily":185,"monthly":168,"yearly":186},1.63,593.93,{"name":188,"vram_gb":166,"cheapest_hour":189,"cost":190,"provider":123,"estimated_tps":194},"RTX 4090",0.1344,{"hourly":189,"daily":191,"monthly":192,"yearly":193},3.23,98.11,1177.34,25200,{"name":196,"vram_gb":166,"cheapest_hour":197,"cost":198,"provider":123,"estimated_tps":194},"RTX 4090D",0.1602,{"hourly":197,"daily":199,"monthly":200,"yearly":201},3.84,116.95,1403.35,{"name":203,"vram_gb":166,"cheapest_hour":204,"cost":205,"provider":123,"estimated_tps":209},"L4",0.2015,{"hourly":204,"daily":206,"monthly":207,"yearly":208},4.84,147.1,1765.14,7500,{"name":211,"vram_gb":166,"cheapest_hour":212,"cost":213,"provider":123,"estimated_tps":217},"A10",0.2409,{"hourly":212,"daily":214,"monthly":215,"yearly":216},5.78,175.86,2110.28,15000,{"posts_count":37,"benchmarks_count":37}]