[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-llama-31-70b-instruct":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":18,"huggingface_url":19,"created_at":20,"seo":21,"category":23,"family":26,"score":28,"variants":29,"quantizations":30,"requirements":57,"gpu_recommendations":142,"fitting_gpus":218,"community":246},19,"llama-31-70b-instruct","Llama-3.1-70B-Instruct","meta-llama\u002FLlama-3.1-70B-Instruct","meta-llama","llama",null,70000000000,"70B",131072,"llama3.1",3100000,6100,"published","70B Llama，全精度需要 2×48GB \u002F 80GB 级别显卡。","https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FLlama-3.1-70B-Instruct","2026-08-12T02:22:24+00:00",{"title":6,"description":18,"image":10,"robots":22,"canonical_url":10},"index, follow",{"name":24,"slug":25},"LLM","llm",{"name":27,"slug":9},"Llama",0.0723,[],[31,36,41,46,50,54],{"format":32,"bits":33,"size_factor":34,"quality_loss":35},"AWQ",4,0.263,0.03,{"format":37,"bits":38,"size_factor":39,"quality_loss":40},"FP16",16,1,0,{"format":42,"bits":43,"size_factor":44,"quality_loss":45},"FP8",8,0.5,0.01,{"format":47,"bits":33,"size_factor":48,"quality_loss":49},"GGUF",0.25,0.05,{"format":51,"bits":33,"size_factor":52,"quality_loss":53},"GPTQ",0.275,0.04,{"format":55,"bits":43,"size_factor":44,"quality_loss":56},"INT8",0.02,{"FP16":58,"FP8":82,"INT8":96,"AWQ":100,"GPTQ":114,"GGUF":128},{"minimum":59,"production":68,"recommended":75},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":62,"kv_cache_gb":44,"vram_gb":63,"ram_gb":64,"disk_gb":65,"requirement_source":66,"confidence":67},"minimum",4096,130.39,169.89,254.83,242.4,"estimated",0.95,{"scenario":69,"context_length":70,"batch_size":33,"weight_gb":62,"kv_cache_gb":71,"vram_gb":72,"ram_gb":73,"disk_gb":74,"requirement_source":66,"confidence":67},"production",16384,64,297.53,446.3,307.4,{"scenario":76,"context_length":77,"batch_size":78,"weight_gb":62,"kv_cache_gb":43,"vram_gb":79,"ram_gb":80,"disk_gb":81,"requirement_source":66,"confidence":67},"recommended",8192,2,193.13,289.7,268.4,{"minimum":83,"production":88,"recommended":92},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":84,"kv_cache_gb":44,"vram_gb":85,"ram_gb":86,"disk_gb":87,"requirement_source":66,"confidence":67},65.19,87.42,131.13,140.7,{"scenario":69,"context_length":70,"batch_size":33,"weight_gb":84,"kv_cache_gb":71,"vram_gb":89,"ram_gb":90,"disk_gb":91,"requirement_source":66,"confidence":67},207.57,311.35,205.7,{"scenario":76,"context_length":77,"batch_size":78,"weight_gb":84,"kv_cache_gb":43,"vram_gb":93,"ram_gb":94,"disk_gb":95,"requirement_source":66,"confidence":67},106.92,160.38,166.7,{"minimum":97,"production":98,"recommended":99},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":84,"kv_cache_gb":44,"vram_gb":85,"ram_gb":86,"disk_gb":87,"requirement_source":66,"confidence":67},{"scenario":69,"context_length":70,"batch_size":33,"weight_gb":84,"kv_cache_gb":71,"vram_gb":89,"ram_gb":90,"disk_gb":91,"requirement_source":66,"confidence":67},{"scenario":76,"context_length":77,"batch_size":78,"weight_gb":84,"kv_cache_gb":43,"vram_gb":93,"ram_gb":94,"disk_gb":95,"requirement_source":66,"confidence":67},{"minimum":101,"production":106,"recommended":110},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":102,"kv_cache_gb":44,"vram_gb":103,"ram_gb":104,"disk_gb":105,"requirement_source":66,"confidence":67},33.82,47.73,71.6,91.76,{"scenario":69,"context_length":70,"batch_size":33,"weight_gb":102,"kv_cache_gb":71,"vram_gb":107,"ram_gb":108,"disk_gb":109,"requirement_source":66,"confidence":67},164.27,246.4,156.76,{"scenario":76,"context_length":77,"batch_size":78,"weight_gb":102,"kv_cache_gb":43,"vram_gb":111,"ram_gb":112,"disk_gb":113,"requirement_source":66,"confidence":67},65.43,98.14,117.76,{"minimum":115,"production":120,"recommended":124},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":116,"kv_cache_gb":44,"vram_gb":117,"ram_gb":118,"disk_gb":119,"requirement_source":66,"confidence":67},34.23,48.25,72.37,92.39,{"scenario":69,"context_length":70,"batch_size":33,"weight_gb":116,"kv_cache_gb":71,"vram_gb":121,"ram_gb":122,"disk_gb":123,"requirement_source":66,"confidence":67},164.83,247.25,157.39,{"scenario":76,"context_length":77,"batch_size":78,"weight_gb":116,"kv_cache_gb":43,"vram_gb":125,"ram_gb":126,"disk_gb":127,"requirement_source":66,"confidence":67},65.96,98.95,118.39,{"minimum":129,"production":134,"recommended":138},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":130,"kv_cache_gb":44,"vram_gb":131,"ram_gb":132,"disk_gb":133,"requirement_source":66,"confidence":67},35.04,49.28,73.92,93.66,{"scenario":69,"context_length":70,"batch_size":33,"weight_gb":130,"kv_cache_gb":71,"vram_gb":135,"ram_gb":136,"disk_gb":137,"requirement_source":66,"confidence":67},165.96,248.93,158.66,{"scenario":76,"context_length":77,"batch_size":78,"weight_gb":130,"kv_cache_gb":43,"vram_gb":139,"ram_gb":140,"disk_gb":141,"requirement_source":66,"confidence":67},67.04,100.56,119.66,{"recommended":143,"cheapest":162,"performance":179,"min_complexity":188,"cluster_options":204,"no_match":215,"required_vram_gb":111,"candidate_count":216,"usage":217},{"gpu_model":144,"gpu_model_id":145,"gpu_count":39,"total_vram_gb":146,"hour_price":147,"monthly_cost":148,"providers":149,"bandwidth_gbps":151,"fp16_tflops":152,"interconnect_score":153,"parallel_efficiency":39,"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",4800,989,40,0.5948,"balanced","tok\u002Fs","141.9 tok\u002Fs",141.9,{"effective_value":160,"effective_unit":156,"capacity_factor":161},99.3,0.7,{"gpu_model":163,"gpu_model_id":164,"gpu_count":43,"total_vram_gb":165,"hour_price":166,"monthly_cost":167,"providers":168,"bandwidth_gbps":170,"fp16_tflops":171,"interconnect_score":153,"parallel_efficiency":172,"score":173,"tier":174,"perf_metric":156,"perf_label":175,"perf_value":176,"workload":10,"cost_metric":177},"RTX 5070",28,96,0.0101,58.98,[169],"vast",672,61,0.4,0.5559,"cheapest","63.6 tok\u002Fs",63.6,{"effective_value":178,"effective_unit":156,"capacity_factor":161},44.5,{"gpu_model":144,"gpu_model_id":145,"gpu_count":43,"total_vram_gb":180,"hour_price":147,"monthly_cost":181,"providers":182,"bandwidth_gbps":151,"fp16_tflops":152,"interconnect_score":153,"parallel_efficiency":172,"score":183,"tier":155,"perf_metric":156,"perf_label":184,"perf_value":185,"workload":10,"cost_metric":186},1128,20440,[150],0.2197,"454.2 tok\u002Fs",454.2,{"effective_value":187,"effective_unit":156,"capacity_factor":161},317.9,{"gpu_model":189,"gpu_model_id":190,"gpu_count":39,"total_vram_gb":191,"hour_price":192,"monthly_cost":193,"providers":194,"bandwidth_gbps":197,"fp16_tflops":198,"interconnect_score":153,"parallel_efficiency":39,"score":199,"tier":155,"perf_metric":156,"perf_label":200,"perf_value":201,"workload":10,"cost_metric":202},"A100 80GB",9,80,1.2,876,[169,195,196,150],"tensordock","runpod",2039,77.97,0.6682,"60.3 tok\u002Fs",60.3,{"effective_value":203,"effective_unit":156,"capacity_factor":161},42.2,[205],{"gpu_model":206,"gpu_model_id":207,"gpu_count":43,"total_vram_gb":71,"hour_price":208,"monthly_cost":209,"providers":210,"bandwidth_gbps":211,"fp16_tflops":207,"interconnect_score":153,"parallel_efficiency":172,"insufficient":212,"vram_shortfall_gb":213,"score":214,"tier":155},"RTX 5060 Ti",24,0.0544,317.7,[169],448,true,1.43,0.5103,false,18,"value",[219,223,231,235,243],{"name":189,"vram_gb":191,"cheapest_hour":192,"cost":220,"provider":169,"estimated_tps":201},{"hourly":192,"daily":221,"monthly":193,"yearly":222},28.8,10512,{"name":224,"vram_gb":191,"cheapest_hour":225,"cost":226,"provider":169,"estimated_tps":230},"H100 SXM",1.6059,{"hourly":225,"daily":227,"monthly":228,"yearly":229},38.54,1172.31,14067.68,99.1,{"name":144,"vram_gb":146,"cheapest_hour":147,"cost":232,"provider":150,"estimated_tps":158},{"hourly":147,"daily":233,"monthly":148,"yearly":234},84,30660,{"name":236,"vram_gb":237,"cheapest_hour":238,"cost":239,"provider":169,"estimated_tps":40},"B200",180,4.1271,{"hourly":238,"daily":240,"monthly":241,"yearly":242},99.05,3012.78,36153.4,{"name":244,"vram_gb":237,"cheapest_hour":10,"cost":10,"provider":10,"estimated_tps":245},"B200 SXM",230.6,{"posts_count":40,"benchmarks_count":40}]