[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"model-qwen3-235b-a22b":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":271,"community":279},9,"qwen3-235b-a22b","Qwen3 235B A22B","Qwen\u002FQwen3-235B-A22B","Qwen","moe_transformer",null,235000000000,"235B",262144,"MIT",460000,3100,"published","MoE 版本，总 235B \u002F 激活 22B，需要多卡或 141GB+ 显存。","https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen3-235B-A22B","2026-08-12T02:22:23+00:00",{"title":6,"description":18,"image":10,"robots":22,"canonical_url":10},"index, follow",{"name":24,"slug":25},"LLM","llm",{"name":8,"slug":27},"qwen",0.1001,[],[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",4096,437.72,558.67,838,721.85,"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",16384,64,721.66,1082.48,786.85,{"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",8192,2,599.59,899.38,747.85,{"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},218.86,281.81,422.71,380.42,{"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},419.63,629.44,445.42,{"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},310.14,465.22,406.42,{"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},113.53,148.57,222.86,216.11,{"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},274.28,411.42,281.11,{"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},170.85,256.27,242.11,{"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},114.9,150.3,225.45,218.25,{"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},276.16,414.25,283.25,{"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},172.66,258.99,244.25,{"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},117.64,153.76,230.64,222.51,{"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},279.94,419.91,287.51,{"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},176.28,264.41,248.51,{"recommended":143,"cheapest":163,"performance":179,"min_complexity":189,"cluster_options":199,"no_match":268,"required_vram_gb":111,"candidate_count":269,"usage":270},{"gpu_model":144,"gpu_model_id":145,"gpu_count":43,"total_vram_gb":146,"hour_price":147,"monthly_cost":148,"providers":149,"bandwidth_gbps":151,"fp16_tflops":152,"interconnect_score":153,"parallel_efficiency":154,"score":155,"tier":156,"perf_metric":157,"perf_label":158,"perf_value":159,"workload":10,"cost_metric":160},"H200 SXM",11,1128,3.5,20440,[150],"lambda",4800,989,40,0.4,0.2601,"balanced","tok\u002Fs","135.3 tok\u002Fs",135.3,{"effective_value":161,"effective_unit":157,"capacity_factor":162},94.7,0.7,{"gpu_model":164,"gpu_model_id":165,"gpu_count":43,"total_vram_gb":166,"hour_price":167,"monthly_cost":168,"providers":169,"bandwidth_gbps":172,"fp16_tflops":173,"interconnect_score":153,"parallel_efficiency":154,"score":174,"tier":156,"perf_metric":157,"perf_label":175,"perf_value":176,"workload":10,"cost_metric":177},"RTX 3090",3,192,0.0678,395.95,[170,171],"vast","tensordock",936,71,0.5359,"26.4 tok\u002Fs",26.4,{"effective_value":178,"effective_unit":157,"capacity_factor":162},18.5,{"gpu_model":144,"gpu_model_id":145,"gpu_count":33,"total_vram_gb":180,"hour_price":147,"monthly_cost":181,"providers":182,"bandwidth_gbps":151,"fp16_tflops":152,"interconnect_score":153,"parallel_efficiency":183,"score":184,"tier":156,"perf_metric":157,"perf_label":185,"perf_value":186,"workload":10,"cost_metric":187},564,10220,[150],0.6,0.3501,"101.5 tok\u002Fs",101.5,{"effective_value":188,"effective_unit":157,"capacity_factor":162},71.1,{"gpu_model":190,"gpu_model_id":191,"gpu_count":39,"total_vram_gb":192,"hour_price":193,"monthly_cost":194,"providers":195,"bandwidth_gbps":40,"fp16_tflops":40,"interconnect_score":153,"parallel_efficiency":39,"score":196,"tier":156,"perf_metric":157,"perf_label":197,"perf_value":40,"workload":10,"cost_metric":198},"B200",27,180,4.1271,3012.78,[170],0.5741,"0 tok\u002Fs",{"effective_value":40,"effective_unit":157,"capacity_factor":162},[200,211,216,225,232,241,248,259],{"gpu_model":201,"gpu_model_id":202,"gpu_count":43,"total_vram_gb":203,"hour_price":204,"monthly_cost":205,"providers":206,"bandwidth_gbps":40,"fp16_tflops":40,"interconnect_score":153,"parallel_efficiency":154,"insufficient":207,"vram_shortfall_gb":208,"score":209,"tier":210},"Tesla V100",17,128,0.0272,158.85,[170],true,42.85,0.5034,"cheapest",{"gpu_model":212,"gpu_model_id":213,"gpu_count":43,"total_vram_gb":203,"hour_price":167,"monthly_cost":168,"providers":214,"bandwidth_gbps":40,"fp16_tflops":40,"interconnect_score":153,"parallel_efficiency":154,"insufficient":207,"vram_shortfall_gb":208,"score":215,"tier":156},"RTX 4070S Ti",22,[170],0.4984,{"gpu_model":217,"gpu_model_id":218,"gpu_count":43,"total_vram_gb":203,"hour_price":219,"monthly_cost":220,"providers":221,"bandwidth_gbps":222,"fp16_tflops":223,"interconnect_score":153,"parallel_efficiency":154,"insufficient":207,"vram_shortfall_gb":208,"score":224,"tier":156},"RTX 5070 Ti",18,0.0849,495.82,[170],896,88,0.5168,{"gpu_model":226,"gpu_model_id":227,"gpu_count":43,"total_vram_gb":203,"hour_price":228,"monthly_cost":229,"providers":230,"bandwidth_gbps":40,"fp16_tflops":40,"interconnect_score":153,"parallel_efficiency":154,"insufficient":207,"vram_shortfall_gb":208,"score":231,"tier":156},"RTX 4080S",21,0.1192,696.13,[170],0.4922,{"gpu_model":233,"gpu_model_id":234,"gpu_count":43,"total_vram_gb":203,"hour_price":235,"monthly_cost":236,"providers":237,"bandwidth_gbps":238,"fp16_tflops":239,"interconnect_score":153,"parallel_efficiency":154,"insufficient":207,"vram_shortfall_gb":208,"score":240,"tier":156},"RTX 5080",23,0.1344,784.9,[170],960,113,0.5123,{"gpu_model":242,"gpu_model_id":243,"gpu_count":43,"total_vram_gb":203,"hour_price":244,"monthly_cost":245,"providers":246,"bandwidth_gbps":40,"fp16_tflops":40,"interconnect_score":153,"parallel_efficiency":154,"insufficient":207,"vram_shortfall_gb":208,"score":247,"tier":156},"RTX PRO 4500",25,0.3222,1881.65,[170],0.4675,{"gpu_model":249,"gpu_model_id":250,"gpu_count":43,"total_vram_gb":251,"hour_price":252,"monthly_cost":253,"providers":254,"bandwidth_gbps":255,"fp16_tflops":256,"interconnect_score":153,"parallel_efficiency":154,"insufficient":207,"vram_shortfall_gb":257,"score":258,"tier":210},"RTX 5070",28,96,0.0101,58.98,[170],672,61,74.85,0.5208,{"gpu_model":260,"gpu_model_id":261,"gpu_count":43,"total_vram_gb":71,"hour_price":262,"monthly_cost":263,"providers":264,"bandwidth_gbps":265,"fp16_tflops":261,"interconnect_score":153,"parallel_efficiency":154,"insufficient":207,"vram_shortfall_gb":266,"score":267,"tier":156},"RTX 5060 Ti",24,0.0544,317.7,[170],448,106.85,0.5103,false,12,"value",[272,276],{"name":190,"vram_gb":192,"cheapest_hour":193,"cost":273,"provider":170,"estimated_tps":40},{"hourly":193,"daily":274,"monthly":194,"yearly":275},99.05,36153.4,{"name":277,"vram_gb":192,"cheapest_hour":10,"cost":10,"provider":10,"estimated_tps":278},"B200 SXM",68.7,{"posts_count":40,"benchmarks_count":40}]