[{"data":1,"prerenderedAt":1251},["ShallowReactive",2],{"header-counts":3,"footer-counts":6,"model-qwen-3":9},{"tools":4,"reviews":5},65,7,{"tools":4,"reviews":5,"playbooks":7,"news":8},10,8,{"id":10,"title":11,"apiCompatible":12,"benchmarks":14,"body":27,"category":1215,"contextWindow":1216,"description":1217,"extension":1218,"maxOutput":1219,"meta":1220,"navigation":325,"path":1221,"pricing":1222,"published":1223,"relatedTools":1224,"releaseDate":1228,"seo":1229,"slug":1230,"stem":1231,"strengths":1232,"updated":1223,"useCases":1238,"vendor":1243,"vendorEn":1244,"weaknesses":1245,"__hash__":1250},"models\u002Fmodels\u002Fqwen-3.md","Qwen 3",[13],"openai",[15,18,21,24],{"name":16,"score":17},"SWE-bench Verified","58.4%",{"name":19,"score":20},"HumanEval","86.7%",{"name":22,"score":23},"MMLU","83.2%",{"name":25,"score":26},"CMMLU","88.9%",{"type":28,"value":29,"toc":1194},"minimark",[30,34,38,41,45,123,131,135,147,274,283,286,289,292,295,299,302,449,453,486,489,561,565,568,615,618,621,726,731,734,780,788,791,885,888,1015,1021,1040,1043,1046,1100,1103,1106,1154,1157,1190],[31,32,33],"h2",{"id":33},"概述",[35,36,37],"p",{},"Qwen 3 是阿里巴巴于 2025 年 9 月发布的第三代通义千问模型系列，覆盖从 0.5B 到 235B 的全尺寸。最大特色是端侧小模型（7B\u002F14B）在同参数级别中性能领先，适合本地部署和移动端使用。",[31,39,40],{"id":40},"核心能力",[42,43,44],"h3",{"id":44},"全尺寸开源",[46,47,48,64],"table",{},[49,50,51],"thead",{},[52,53,54,58,61],"tr",{},[55,56,57],"th",{},"参数量",[55,59,60],{},"适用场景",[55,62,63],{},"显存需求",[65,66,67,79,90,101,112],"tbody",{},[52,68,69,73,76],{},[70,71,72],"td",{},"0.5B \u002F 1.5B",[70,74,75],{},"手机端、IoT、浏览器 WASM",[70,77,78],{},"\u003C 2GB",[52,80,81,84,87],{},[70,82,83],{},"7B \u002F 14B",[70,85,86],{},"本地开发、个人使用",[70,88,89],{},"8-16GB",[52,91,92,95,98],{},[70,93,94],{},"32B \u002F 72B",[70,96,97],{},"企业部署",[70,99,100],{},"48-128GB",[52,102,103,106,109],{},[70,104,105],{},"Qwen3-MoE-A14B-72B",[70,107,108],{},"性价比之选",[70,110,111],{},"64-96GB",[52,113,114,117,120],{},[70,115,116],{},"235B",[70,118,119],{},"云端 API",[70,121,122],{},"多卡 H100",[35,124,125,126,130],{},"端侧 7B 模型在 MMLU 上达到 75%+，足以胜任日常对话、简单编程、文档摘要等任务。",[127,128,129],"strong",{},"这是 Qwen 系列最大的差异化优势","——你能拿 7B 跑出接近 GPT-3.5 体验的本地模型，对隐私敏感的个人和企业是刚需。",[42,132,134],{"id":133},"hybrid-thinking混合推理模式","Hybrid Thinking（混合推理模式）",[35,136,137,138,141,142,146],{},"Qwen 3 引入了",[127,139,140],{},"可切换的推理模式","——同一个模型，加 ",[143,144,145],"code",{},"enable_thinking=True"," 就变推理模型：",[148,149,154],"pre",{"className":150,"code":151,"language":152,"meta":153,"style":153},"language-python shiki shiki-themes github-light github-dark","resp = client.chat.completions.create(\n    model=\"qwen3-235b-a22b\",\n    messages=[{\"role\": \"user\", \"content\": \"证明...\"}],\n    extra_body={\n        \"enable_thinking\": True,    # 开 thinking\n    },\n)\n# resp.choices[0].message.reasoning_content 是思维链\n# resp.choices[0].message.content 是最终答案\n","python","",[143,155,156,172,188,222,233,252,258,263,268],{"__ignoreMap":153},[157,158,161,165,169],"span",{"class":159,"line":160},"line",1,[157,162,164],{"class":163},"sVt8B","resp ",[157,166,168],{"class":167},"szBVR","=",[157,170,171],{"class":163}," client.chat.completions.create(\n",[157,173,175,179,181,185],{"class":159,"line":174},2,[157,176,178],{"class":177},"s4XuR","    model",[157,180,168],{"class":167},[157,182,184],{"class":183},"sZZnC","\"qwen3-235b-a22b\"",[157,186,187],{"class":163},",\n",[157,189,191,194,196,199,202,205,208,211,214,216,219],{"class":159,"line":190},3,[157,192,193],{"class":177},"    messages",[157,195,168],{"class":167},[157,197,198],{"class":163},"[{",[157,200,201],{"class":183},"\"role\"",[157,203,204],{"class":163},": ",[157,206,207],{"class":183},"\"user\"",[157,209,210],{"class":163},", ",[157,212,213],{"class":183},"\"content\"",[157,215,204],{"class":163},[157,217,218],{"class":183},"\"证明...\"",[157,220,221],{"class":163},"}],\n",[157,223,225,228,230],{"class":159,"line":224},4,[157,226,227],{"class":177},"    extra_body",[157,229,168],{"class":167},[157,231,232],{"class":163},"{\n",[157,234,236,239,241,245,248],{"class":159,"line":235},5,[157,237,238],{"class":183},"        \"enable_thinking\"",[157,240,204],{"class":163},[157,242,244],{"class":243},"sj4cs","True",[157,246,247],{"class":163},",    ",[157,249,251],{"class":250},"sJ8bj","# 开 thinking\n",[157,253,255],{"class":159,"line":254},6,[157,256,257],{"class":163},"    },\n",[157,259,260],{"class":159,"line":5},[157,261,262],{"class":163},")\n",[157,264,265],{"class":159,"line":8},[157,266,267],{"class":250},"# resp.choices[0].message.reasoning_content 是思维链\n",[157,269,271],{"class":159,"line":270},9,[157,272,273],{"class":250},"# resp.choices[0].message.content 是最终答案\n",[35,275,276,277,282],{},"对比 ",[278,279,281],"a",{"href":280},"\u002Fmodels\u002Fdeepseek-r1.html","DeepSeek-R1"," 那种\"必须用单独模型\"，Qwen 3 一个模型双模式更省事——简单任务关 thinking 快、复杂题开 thinking 强。",[42,284,285],{"id":285},"多语言",[35,287,288],{},"中英日韩阿法德七种语言均衡发展，不像某些模型偏科中文或英文。对于需要多语言支持的应用（跨境电商、国际客服），Qwen 3 是理想选择。覆盖语种数 119 种，业界最广。",[42,290,291],{"id":291},"阿里云生态",[35,293,294],{},"通过阿里云百炼平台可以直接调用，与其他阿里云服务（OSS、RDS、函数计算）集成方便。免费额度：个人开发者每月 200 万 token。",[31,296,298],{"id":297},"api-调用示例","API 调用示例",[42,300,301],{"id":301},"阿里云百炼平台",[148,303,305],{"className":150,"code":304,"language":152,"meta":153,"style":153},"from openai import OpenAI\n\nclient = OpenAI(\n    api_key=\"sk-...\",\n    base_url=\"https:\u002F\u002Fdashscope.aliyuncs.com\u002Fcompatible-mode\u002Fv1\",\n)\n\nresp = client.chat.completions.create(\n    model=\"qwen-plus\",       # 旗舰别名（背后是 Qwen3-235B）\n    temperature=0.3,\n    messages=[\n        {\"role\": \"user\", \"content\": \"...\"},\n    ],\n)\n",[143,306,307,321,327,337,349,361,365,369,377,392,404,414,438,444],{"__ignoreMap":153},[157,308,309,312,315,318],{"class":159,"line":160},[157,310,311],{"class":167},"from",[157,313,314],{"class":163}," openai ",[157,316,317],{"class":167},"import",[157,319,320],{"class":163}," OpenAI\n",[157,322,323],{"class":159,"line":174},[157,324,326],{"emptyLinePlaceholder":325},true,"\n",[157,328,329,332,334],{"class":159,"line":190},[157,330,331],{"class":163},"client ",[157,333,168],{"class":167},[157,335,336],{"class":163}," OpenAI(\n",[157,338,339,342,344,347],{"class":159,"line":224},[157,340,341],{"class":177},"    api_key",[157,343,168],{"class":167},[157,345,346],{"class":183},"\"sk-...\"",[157,348,187],{"class":163},[157,350,351,354,356,359],{"class":159,"line":235},[157,352,353],{"class":177},"    base_url",[157,355,168],{"class":167},[157,357,358],{"class":183},"\"https:\u002F\u002Fdashscope.aliyuncs.com\u002Fcompatible-mode\u002Fv1\"",[157,360,187],{"class":163},[157,362,363],{"class":159,"line":254},[157,364,262],{"class":163},[157,366,367],{"class":159,"line":5},[157,368,326],{"emptyLinePlaceholder":325},[157,370,371,373,375],{"class":159,"line":8},[157,372,164],{"class":163},[157,374,168],{"class":167},[157,376,171],{"class":163},[157,378,379,381,383,386,389],{"class":159,"line":270},[157,380,178],{"class":177},[157,382,168],{"class":167},[157,384,385],{"class":183},"\"qwen-plus\"",[157,387,388],{"class":163},",       ",[157,390,391],{"class":250},"# 旗舰别名（背后是 Qwen3-235B）\n",[157,393,394,397,399,402],{"class":159,"line":7},[157,395,396],{"class":177},"    temperature",[157,398,168],{"class":167},[157,400,401],{"class":243},"0.3",[157,403,187],{"class":163},[157,405,407,409,411],{"class":159,"line":406},11,[157,408,193],{"class":177},[157,410,168],{"class":167},[157,412,413],{"class":163},"[\n",[157,415,417,420,422,424,426,428,430,432,435],{"class":159,"line":416},12,[157,418,419],{"class":163},"        {",[157,421,201],{"class":183},[157,423,204],{"class":163},[157,425,207],{"class":183},[157,427,210],{"class":163},[157,429,213],{"class":183},[157,431,204],{"class":163},[157,433,434],{"class":183},"\"...\"",[157,436,437],{"class":163},"},\n",[157,439,441],{"class":159,"line":440},13,[157,442,443],{"class":163},"    ],\n",[157,445,447],{"class":159,"line":446},14,[157,448,262],{"class":163},[42,450,452],{"id":451},"本地-ollama-部署无需联网","本地 Ollama 部署（无需联网）",[148,454,458],{"className":455,"code":456,"language":457,"meta":153,"style":153},"language-bash shiki shiki-themes github-light github-dark","# 一键拉取 7B 模型，笔记本能跑\nollama pull qwen3:7b\nollama run qwen3:7b\n","bash",[143,459,460,465,477],{"__ignoreMap":153},[157,461,462],{"class":159,"line":160},[157,463,464],{"class":250},"# 一键拉取 7B 模型，笔记本能跑\n",[157,466,467,471,474],{"class":159,"line":174},[157,468,470],{"class":469},"sScJk","ollama",[157,472,473],{"class":183}," pull",[157,475,476],{"class":183}," qwen3:7b\n",[157,478,479,481,484],{"class":159,"line":190},[157,480,470],{"class":469},[157,482,483],{"class":183}," run",[157,485,476],{"class":183},[35,487,488],{},"然后通过 OpenAI 兼容 API 接入任何工具：",[148,490,492],{"className":150,"code":491,"language":152,"meta":153,"style":153},"client = OpenAI(\n    api_key=\"ollama\",\n    base_url=\"http:\u002F\u002Flocalhost:11434\u002Fv1\",\n)\nresp = client.chat.completions.create(model=\"qwen3:7b\", messages=[...])\n",[143,493,494,502,513,524,528],{"__ignoreMap":153},[157,495,496,498,500],{"class":159,"line":160},[157,497,331],{"class":163},[157,499,168],{"class":167},[157,501,336],{"class":163},[157,503,504,506,508,511],{"class":159,"line":174},[157,505,341],{"class":177},[157,507,168],{"class":167},[157,509,510],{"class":183},"\"ollama\"",[157,512,187],{"class":163},[157,514,515,517,519,522],{"class":159,"line":190},[157,516,353],{"class":177},[157,518,168],{"class":167},[157,520,521],{"class":183},"\"http:\u002F\u002Flocalhost:11434\u002Fv1\"",[157,523,187],{"class":163},[157,525,526],{"class":159,"line":224},[157,527,262],{"class":163},[157,529,530,532,534,537,540,542,545,547,550,552,555,558],{"class":159,"line":235},[157,531,164],{"class":163},[157,533,168],{"class":167},[157,535,536],{"class":163}," client.chat.completions.create(",[157,538,539],{"class":177},"model",[157,541,168],{"class":167},[157,543,544],{"class":183},"\"qwen3:7b\"",[157,546,210],{"class":163},[157,548,549],{"class":177},"messages",[157,551,168],{"class":167},[157,553,554],{"class":163},"[",[157,556,557],{"class":243},"...",[157,559,560],{"class":163},"])\n",[42,562,564],{"id":563},"vllm-高性能服务","vLLM 高性能服务",[35,566,567],{},"企业部署推荐 vLLM：",[148,569,571],{"className":455,"code":570,"language":457,"meta":153,"style":153},"vllm serve Qwen\u002FQwen3-72B \\\n    --tensor-parallel-size 4 \\\n    --max-model-len 131072 \\\n    --enable-prefix-caching   # 开 prompt cache\n",[143,572,573,587,597,607],{"__ignoreMap":153},[157,574,575,578,581,584],{"class":159,"line":160},[157,576,577],{"class":469},"vllm",[157,579,580],{"class":183}," serve",[157,582,583],{"class":183}," Qwen\u002FQwen3-72B",[157,585,586],{"class":243}," \\\n",[157,588,589,592,595],{"class":159,"line":174},[157,590,591],{"class":243},"    --tensor-parallel-size",[157,593,594],{"class":243}," 4",[157,596,586],{"class":243},[157,598,599,602,605],{"class":159,"line":190},[157,600,601],{"class":243},"    --max-model-len",[157,603,604],{"class":243}," 131072",[157,606,586],{"class":243},[157,608,609,612],{"class":159,"line":224},[157,610,611],{"class":243},"    --enable-prefix-caching",[157,613,614],{"class":250},"   # 开 prompt cache\n",[35,616,617],{},"实测 4×A100 跑 72B 单并发 30 tok\u002Fs，10 并发总吞吐 200+ tok\u002Fs——比走 API 长期看更省钱。",[31,619,620],{"id":620},"模型家族",[46,622,623,642],{},[49,624,625],{},[52,626,627,630,633,636,639],{},[55,628,629],{},"别名（百炼）",[55,631,632],{},"实际型号",[55,634,635],{},"Input",[55,637,638],{},"Output",[55,640,641],{},"上下文",[65,643,644,661,677,694,710],{},[52,645,646,649,652,655,658],{},[70,647,648],{},"qwen-turbo",[70,650,651],{},"Qwen3-30B-MoE",[70,653,654],{},"¥0.3\u002FM",[70,656,657],{},"¥0.6\u002FM",[70,659,660],{},"128K",[52,662,663,666,669,672,675],{},[70,664,665],{},"qwen-plus",[70,667,668],{},"Qwen3-72B \u002F 235B",[70,670,671],{},"¥0.8\u002FM",[70,673,674],{},"¥2\u002FM",[70,676,660],{},[52,678,679,682,685,688,691],{},[70,680,681],{},"qwen-max",[70,683,684],{},"Qwen3-Max",[70,686,687],{},"¥2.4\u002FM",[70,689,690],{},"¥9.6\u002FM",[70,692,693],{},"32K",[52,695,696,699,702,705,707],{},[70,697,698],{},"qwen-long",[70,700,701],{},"Qwen-Long",[70,703,704],{},"¥0.5\u002FM",[70,706,674],{},[70,708,709],{},"10M",[52,711,712,715,718,721,724],{},[70,713,714],{},"qwen-vl-plus",[70,716,717],{},"Qwen3-VL",[70,719,720],{},"¥1.5\u002FM",[70,722,723],{},"¥4.5\u002FM",[70,725,693],{},[35,727,728,730],{},[143,729,698],{}," 上下文 1000 万 token——业界最长，专门做超长文档场景。",[31,732,733],{"id":733},"定价",[46,735,736,746],{},[49,737,738],{},[52,739,740,743],{},[55,741,742],{},"项目",[55,744,745],{},"价格",[65,747,748,756,764,772],{},[52,749,750,753],{},[70,751,752],{},"Input (qwen-plus)",[70,754,755],{},"¥0.8 \u002F 百万 token",[52,757,758,761],{},[70,759,760],{},"Output (qwen-plus)",[70,762,763],{},"¥2 \u002F 百万 token",[52,765,766,769],{},[70,767,768],{},"免费额度",[70,770,771],{},"每月 200 万 token",[52,773,774,777],{},[70,775,776],{},"Batch API",[70,778,779],{},"-50%",[35,781,782,783,787],{},"是 ",[278,784,786],{"href":785},"\u002Fmodels\u002Fdeepseek-v3.html","DeepSeek-V3"," 之后第二便宜的选项。",[31,789,790],{"id":790},"关键参数",[46,792,793,806],{},[49,794,795],{},[52,796,797,800,803],{},[55,798,799],{},"参数",[55,801,802],{},"推荐",[55,804,805],{},"说明",[65,807,808,821,833,846,859,872],{},[52,809,810,815,818],{},[70,811,812],{},[143,813,814],{},"temperature",[70,816,817],{},"0.7（默认）",[70,819,820],{},"通用",[52,822,823,827,830],{},[70,824,825],{},[143,826,814],{},[70,828,829],{},"0",[70,831,832],{},"工具调用 \u002F 代码",[52,834,835,840,843],{},[70,836,837],{},[143,838,839],{},"top_p",[70,841,842],{},"0.8",[70,844,845],{},"默认",[52,847,848,853,856],{},[70,849,850],{},[143,851,852],{},"max_tokens",[70,854,855],{},"显式",[70,857,858],{},"默认 1500 太短",[52,860,861,866,869],{},[70,862,863],{},[143,864,865],{},"enable_thinking",[70,867,868],{},"False \u002F True",[70,870,871],{},"是否开推理模式",[52,873,874,879,882],{},[70,875,876],{},[143,877,878],{},"repetition_penalty",[70,880,881],{},"1.05",[70,883,884],{},"防止重复",[31,886,887],{"id":887},"与同档国产模型怎么选",[46,889,890,907],{},[49,891,892],{},[52,893,894,897,899,902,904],{},[55,895,896],{},"维度",[55,898,11],{},[55,900,901],{},"GLM-5.2",[55,903,786],{},[55,905,906],{},"Kimi K2",[65,908,909,924,937,952,967,982,998],{},[52,910,911,914,917,920,922],{},[70,912,913],{},"编程",[70,915,916],{},"★★★☆☆",[70,918,919],{},"★★★★☆",[70,921,919],{},[70,923,916],{},[52,925,926,928,931,933,935],{},[70,927,285],{},[70,929,930],{},"★★★★★",[70,932,916],{},[70,934,916],{},[70,936,916],{},[52,938,939,942,944,946,949],{},[70,940,941],{},"端侧部署",[70,943,930],{},[70,945,916],{},[70,947,948],{},"★☆☆☆☆",[70,950,951],{},"❌",[52,953,954,957,960,962,965],{},[70,955,956],{},"Hybrid Thinking",[70,958,959],{},"✅",[70,961,951],{},[70,963,964],{},"分别用 R1",[70,966,951],{},[52,968,969,972,974,976,979],{},[70,970,971],{},"价格 Input",[70,973,671],{},[70,975,674],{},[70,977,978],{},"¥1\u002FM",[70,980,981],{},"¥4\u002FM",[52,983,984,987,990,993,996],{},[70,985,986],{},"开源",[70,988,989],{},"✅ 全系列",[70,991,992],{},"部分",[70,994,995],{},"✅ MoE",[70,997,951],{},[52,999,1000,1003,1006,1009,1012],{},[70,1001,1002],{},"国内云生态",[70,1004,1005],{},"阿里云全套",[70,1007,1008],{},"智谱独立",[70,1010,1011],{},"独立",[70,1013,1014],{},"Moonshot",[35,1016,1017,1020],{},[127,1018,1019],{},"建议","：",[1022,1023,1024,1028,1031,1034,1037],"ul",{},[1025,1026,1027],"li",{},"本地部署 \u002F 端侧 → Qwen 3（7B 是同档最强）",[1025,1029,1030],{},"多语言应用 → Qwen 3",[1025,1032,1033],{},"主力编程 → GLM-5.2",[1025,1035,1036],{},"极致低成本 → DeepSeek-V3",[1025,1038,1039],{},"长文档分析 → Kimi K2 \u002F qwen-long",[31,1041,1042],{"id":1042},"端侧场景的真实价值",[35,1044,1045],{},"7B 模型本地能跑意味着：",[46,1047,1048,1058],{},[49,1049,1050],{},[52,1051,1052,1055],{},[55,1053,1054],{},"场景",[55,1056,1057],{},"价值",[65,1059,1060,1068,1076,1084,1092],{},[52,1061,1062,1065],{},[70,1063,1064],{},"个人助手",[70,1066,1067],{},"笔记\u002F邮件\u002F搜索本地处理，数据不出本机",[52,1069,1070,1073],{},[70,1071,1072],{},"IoT 设备",[70,1074,1075],{},"离线语音对话、智能音箱",[52,1077,1078,1081],{},[70,1079,1080],{},"浏览器扩展",[70,1082,1083],{},"离线网页摘要 \u002F 翻译",[52,1085,1086,1089],{},[70,1087,1088],{},"移动 App",[70,1090,1091],{},"iOS\u002FAndroid 端侧推理",[52,1093,1094,1097],{},[70,1095,1096],{},"企业边缘",[70,1098,1099],{},"工厂 \u002F 医院本地部署",[35,1101,1102],{},"Qwen 3 是国内做这类场景的事实标准。",[31,1104,1105],{"id":1105},"避坑清单",[1022,1107,1108,1120,1128,1134,1140,1148],{},[1025,1109,1110,1115,1116,1119],{},[127,1111,1112,1114],{},[143,1113,665],{}," 别名背后型号变化","：阿里偶尔切换后台模型，行为可能小幅变化。生产用 ",[143,1117,1118],{},"qwen3-235b-a22b"," 这种带版本的具体型号。",[1025,1121,1122,1127],{},[127,1123,1124,1126],{},[143,1125,145],{}," 不要白开","：简单任务开了变慢且更贵，按需启用。",[1025,1129,1130,1133],{},[127,1131,1132],{},"多语言不等于翻译","：跨语言生成可能掺杂代码切换问题，重要场景仍建议用专业翻译模型。",[1025,1135,1136,1139],{},[127,1137,1138],{},"本地 Ollama 量化精度","：默认 q4 量化质量损失明显，重要场景用 q8 或 fp16。",[1025,1141,1142,1147],{},[127,1143,1144,1146],{},[143,1145,698],{}," 价格陷阱","：上下文越长单次成本越高，10M 上下文用一次几十块。",[1025,1149,1150,1153],{},[127,1151,1152],{},"Batch API 24h 出结果","：实时任务别走 batch。",[31,1155,1156],{"id":1156},"延伸阅读",[1022,1158,1159,1167,1174,1183],{},[1025,1160,1161,1162,1166],{},"本地部署：",[278,1163,1165],{"href":1164},"\u002Fwiki\u002Flora.html","LoRA","（在 Qwen 上微调）",[1025,1168,1169,1170],{},"推理模式原理：",[278,1171,1173],{"href":1172},"\u002Fwiki\u002Ftemperature-top-p.html","Temperature 与 Top-P",[1025,1175,1176,1177,1179,1180],{},"同档对比：",[278,1178,786],{"href":785}," \u002F ",[278,1181,901],{"href":1182},"\u002Fmodels\u002Fglm-5.2.html",[1025,1184,1185,1186],{},"多语言场景：",[278,1187,1189],{"href":1188},"\u002Fwiki\u002Fembedding.html","Embedding",[1191,1192,1193],"style",{},"html pre.shiki code .sVt8B, html code.shiki .sVt8B{--shiki-default:#24292E;--shiki-dark:#E1E4E8}html pre.shiki code .szBVR, html code.shiki .szBVR{--shiki-default:#D73A49;--shiki-dark:#F97583}html pre.shiki code .s4XuR, html code.shiki .s4XuR{--shiki-default:#E36209;--shiki-dark:#FFAB70}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html pre.shiki code .sJ8bj, html code.shiki .sJ8bj{--shiki-default:#6A737D;--shiki-dark:#6A737D}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}",{"title":153,"searchDepth":190,"depth":190,"links":1195},[1196,1197,1203,1208,1209,1210,1211,1212,1213,1214],{"id":33,"depth":174,"text":33},{"id":40,"depth":174,"text":40,"children":1198},[1199,1200,1201,1202],{"id":44,"depth":190,"text":44},{"id":133,"depth":190,"text":134},{"id":285,"depth":190,"text":285},{"id":291,"depth":190,"text":291},{"id":297,"depth":174,"text":298,"children":1204},[1205,1206,1207],{"id":301,"depth":190,"text":301},{"id":451,"depth":190,"text":452},{"id":563,"depth":190,"text":564},{"id":620,"depth":174,"text":620},{"id":733,"depth":174,"text":733},{"id":790,"depth":174,"text":790},{"id":887,"depth":174,"text":887},{"id":1042,"depth":174,"text":1042},{"id":1105,"depth":174,"text":1105},{"id":1156,"depth":174,"text":1156},"llm",131072,"阿里通义千问 Qwen 3 系列开源大模型，0.5B 至 235B 全尺寸覆盖，端侧小模型性能业界领先，多语言（119 种）能力广泛，国内可直连 + 慷慨免费 API 额度，适合从边缘设备到云端 GPU 的全栈部署。","md",16384,{},"\u002Fmodels\u002Fqwen-3","Input ¥0.8\u002FM · Output ¥2\u002FM（免费 API 额度）","2026-06-21",[1225,1226,1227],"coding\u002Flocal\u002Follama","coding\u002Flocal\u002Flm-studio","coding\u002Fapi\u002Fopenrouter","2025-09-25",{"title":11,"description":1217},"qwen-3","models\u002Fqwen-3",[1233,1234,1235,1236,1237],"开源全系列（0.5B 到 235B），覆盖端侧到云端","多语言能力强，中英日韩阿法德表现均衡","免费 API 额度大方，阿里云百炼平台可直接调用","端侧小模型（7B\u002F14B）性能领先，适合本地部署","支持长文档理解和代码生成",[1239,1240,1241,1242],"端侧 AI 部署（手机\u002FIoT\u002F本地）","多语言应用开发","低成本 API 调用","企业私有化部署","阿里巴巴","Alibaba",[1246,1247,1248,1249],"编程实操不如 Claude Sonnet 4 \u002F GLM-5.2","Agent 工具调用生态不如 Anthropic 成熟","16K 输出窗口偏短","多步推理稳定性一般","sRIwsG5ksoXekmio_f5Yx0BZ-xFnNldjlIysLnb_h74",1782316489332]