[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-gemma-3-12b":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":29,"variants":30,"quantizations":31,"requirements":57,"gpu_recommendations":139,"fitting_gpus":206,"community":241},16,"gemma-3-12b","Gemma 3 12B","google\u002Fgemma-3-12b-it","google","transformer",null,12000000000,"12B",131072,"MIT",520000,2100,"published","Gemma 3 12B, fits single consumer GPU.","https:\u002F\u002Fhuggingface.co\u002Fgoogle\u002Fgemma-3-12b-it","2026-08-12T02:22:24+00:00",{"title":6,"description":18,"image":10,"robots":22,"canonical_url":10},"index, follow",{"name":24,"slug":25},"OCR \u002F Vision","ocr-vision",{"name":27,"slug":28},"Gemma","gemma",0.274,[],[32,37,41,46,50,54],{"format":33,"bits":34,"size_factor":35,"quality_loss":36},"AWQ",4,0.263,0.03,{"format":38,"bits":4,"size_factor":39,"quality_loss":40},"FP16",1,0,{"format":42,"bits":43,"size_factor":44,"quality_loss":45},"FP8",8,0.5,0.01,{"format":47,"bits":34,"size_factor":48,"quality_loss":49},"GGUF",0.25,0.05,{"format":51,"bits":34,"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":113,"GGUF":126},{"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",4096,22.35,33.22,49.84,73.87,"estimated",0.95,{"scenario":69,"context_length":70,"batch_size":34,"weight_gb":62,"kv_cache_gb":71,"vram_gb":72,"ram_gb":73,"disk_gb":74,"requirement_source":66,"confidence":67},"production",16384,64,148.45,222.67,138.87,{"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",8192,2,50.26,75.39,99.87,{"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},11.18,19.09,32,56.43,{"scenario":69,"context_length":70,"batch_size":34,"weight_gb":84,"kv_cache_gb":71,"vram_gb":89,"ram_gb":90,"disk_gb":91,"requirement_source":66,"confidence":67},133.02,199.53,121.43,{"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},35.48,53.22,82.43,{"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":34,"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":105,"recommended":109},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":102,"kv_cache_gb":44,"vram_gb":103,"ram_gb":86,"disk_gb":104,"requirement_source":66,"confidence":67},5.8,12.28,48.04,{"scenario":69,"context_length":70,"batch_size":34,"weight_gb":102,"kv_cache_gb":71,"vram_gb":106,"ram_gb":107,"disk_gb":108,"requirement_source":66,"confidence":67},125.6,188.4,113.04,{"scenario":76,"context_length":77,"batch_size":78,"weight_gb":102,"kv_cache_gb":43,"vram_gb":110,"ram_gb":111,"disk_gb":112,"requirement_source":66,"confidence":67},28.37,42.55,74.04,{"minimum":114,"production":118,"recommended":122},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":115,"kv_cache_gb":44,"vram_gb":116,"ram_gb":86,"disk_gb":117,"requirement_source":66,"confidence":67},5.87,12.37,48.15,{"scenario":69,"context_length":70,"batch_size":34,"weight_gb":115,"kv_cache_gb":71,"vram_gb":119,"ram_gb":120,"disk_gb":121,"requirement_source":66,"confidence":67},125.7,188.55,113.15,{"scenario":76,"context_length":77,"batch_size":78,"weight_gb":115,"kv_cache_gb":43,"vram_gb":123,"ram_gb":124,"disk_gb":125,"requirement_source":66,"confidence":67},28.46,42.69,74.15,{"minimum":127,"production":131,"recommended":135},{"scenario":60,"context_length":61,"batch_size":39,"weight_gb":128,"kv_cache_gb":44,"vram_gb":129,"ram_gb":86,"disk_gb":130,"requirement_source":66,"confidence":67},6.01,12.55,48.37,{"scenario":69,"context_length":70,"batch_size":34,"weight_gb":128,"kv_cache_gb":71,"vram_gb":132,"ram_gb":133,"disk_gb":134,"requirement_source":66,"confidence":67},125.89,188.83,113.37,{"scenario":76,"context_length":77,"batch_size":78,"weight_gb":128,"kv_cache_gb":43,"vram_gb":136,"ram_gb":137,"disk_gb":138,"requirement_source":66,"confidence":67},28.64,42.97,74.37,{"recommended":140,"cheapest":162,"performance":173,"min_complexity":188,"cluster_options":202,"no_match":203,"required_vram_gb":110,"candidate_count":204,"usage":205},{"gpu_model":141,"gpu_model_id":142,"gpu_count":39,"total_vram_gb":143,"hour_price":144,"monthly_cost":145,"providers":146,"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},"H100 SXM",10,80,1.6022,1169.61,[147,148,149,150],"vast","tensordock","lambda","runpod",3350,66.9,40,0.5841,"balanced","tok\u002Fs","577.6 tok\u002Fs",577.6,{"effective_value":160,"effective_unit":156,"capacity_factor":161},404.3,0.7,{"gpu_model":163,"gpu_model_id":164,"gpu_count":78,"total_vram_gb":86,"hour_price":165,"monthly_cost":166,"providers":167,"bandwidth_gbps":40,"fp16_tflops":40,"interconnect_score":153,"parallel_efficiency":168,"score":169,"tier":170,"perf_metric":156,"perf_label":171,"perf_value":40,"workload":10,"cost_metric":172},"Tesla V100",17,0.0272,39.71,[147],0.8,0.6103,"cheapest","0 tok\u002Fs",{"effective_value":40,"effective_unit":156,"capacity_factor":161},{"gpu_model":174,"gpu_model_id":175,"gpu_count":43,"total_vram_gb":176,"hour_price":177,"monthly_cost":178,"providers":179,"bandwidth_gbps":180,"fp16_tflops":181,"interconnect_score":153,"parallel_efficiency":182,"score":183,"tier":155,"perf_metric":156,"perf_label":184,"perf_value":185,"workload":10,"cost_metric":186},"H200 SXM",11,1128,3.5,20440,[149],4800,989,0.4,0.2197,"2648.3 tok\u002Fs",2648.3,{"effective_value":187,"effective_unit":156,"capacity_factor":161},1853.8,{"gpu_model":189,"gpu_model_id":190,"gpu_count":39,"total_vram_gb":191,"hour_price":192,"monthly_cost":193,"providers":194,"bandwidth_gbps":195,"fp16_tflops":196,"interconnect_score":153,"parallel_efficiency":39,"score":197,"tier":170,"perf_metric":156,"perf_label":198,"perf_value":199,"workload":10,"cost_metric":200},"RTX A6000",6,48,0.3001,219.07,[147,149],768,77,0.6485,"132.4 tok\u002Fs",132.4,{"effective_value":201,"effective_unit":156,"capacity_factor":161},92.7,[],false,24,"value",[207,211,219,227,234],{"name":189,"vram_gb":191,"cheapest_hour":192,"cost":208,"provider":147,"estimated_tps":199},{"hourly":192,"daily":209,"monthly":193,"yearly":210},7.2,2628.88,{"name":212,"vram_gb":86,"cheapest_hour":213,"cost":214,"provider":147,"estimated_tps":218},"RTX 5090",0.3222,{"hourly":213,"daily":215,"monthly":216,"yearly":217},7.73,235.21,2822.47,309,{"name":220,"vram_gb":191,"cheapest_hour":221,"cost":222,"provider":147,"estimated_tps":226},"L40S",0.4669,{"hourly":221,"daily":223,"monthly":224,"yearly":225},11.21,340.84,4090.04,149,{"name":228,"vram_gb":191,"cheapest_hour":229,"cost":230,"provider":147,"estimated_tps":40},"RTX PRO 6000 WS",0.8022,{"hourly":229,"daily":231,"monthly":232,"yearly":233},19.25,585.61,7027.27,{"name":235,"vram_gb":191,"cheapest_hour":236,"cost":237,"provider":147,"estimated_tps":40},"RTX PRO 6000 S",1.0131,{"hourly":236,"daily":238,"monthly":239,"yearly":240},24.31,739.56,8874.76,{"posts_count":40,"benchmarks_count":40}]