{"id":722,"date":"2026-07-26T09:09:54","date_gmt":"2026-07-26T01:09:54","guid":{"rendered":"https:\/\/book.yuekegu.com\/index.php\/2026\/07\/26\/%e9%ad%94%e6%90%ad%e7%a4%be%e5%8c%ba%ef%bc%9a%e6%af%8f%e5%a4%a9%e7%99%bd%e5%ab%962000%e6%ac%a1ai%e5%a4%a7%e6%a8%a1%e5%9e%8bapi-%e5%bc%80%e5%8f%91%e8%80%85%e7%9a%84%e5%85%8d%e8%b4%b9\/"},"modified":"2026-07-26T09:09:54","modified_gmt":"2026-07-26T01:09:54","slug":"%e9%ad%94%e6%90%ad%e7%a4%be%e5%8c%ba%ef%bc%9a%e6%af%8f%e5%a4%a9%e7%99%bd%e5%ab%962000%e6%ac%a1ai%e5%a4%a7%e6%a8%a1%e5%9e%8bapi-%e5%bc%80%e5%8f%91%e8%80%85%e7%9a%84%e5%85%8d%e8%b4%b9","status":"publish","type":"post","link":"https:\/\/book.yuekegu.com\/index.php\/2026\/07\/26\/%e9%ad%94%e6%90%ad%e7%a4%be%e5%8c%ba%ef%bc%9a%e6%af%8f%e5%a4%a9%e7%99%bd%e5%ab%962000%e6%ac%a1ai%e5%a4%a7%e6%a8%a1%e5%9e%8bapi-%e5%bc%80%e5%8f%91%e8%80%85%e7%9a%84%e5%85%8d%e8%b4%b9\/","title":{"rendered":"\u9b54\u642d\u793e\u533a\uff1a\u6bcf\u5929\u767d\u5ad62000\u6b21AI\u5927\u6a21\u578bAPI\u2014\u2014\u5f00\u53d1\u8005\u7684\u514d\u8d39\u5f39\u836f\u5e93"},"content":{"rendered":"\n

\u672c\u6587\u8f6c\u8f7d\u81ea AITimes \/ \u5fae\u4fe1\u516c\u4f17\u53f7\uff0c\u539f\u6587\u4f5c\u8005\uff1a\u6797\u540c\u5b66 | AI \u5f00\u53d1\u5b9e\u6218\u7cfb\u5217\uff0c\u539f\u6587\u94fe\u63a5\uff1ahttps:\/\/mp.weixin.qq.com\/s\/rm_38Tz8qjsS_qcuWY2Sqw<\/a><\/p>\n\n\n\n

\ud83d\udcb0 \u7701\u94b1\u624d\u662f\u786c\u9053\u7406<\/h2>\n\n\n\n

\u641e AI \u5f00\u53d1\uff0c\u6a21\u578b\u8c03\u7528\u8d39\u662f\u7ed5\u4e0d\u5f00\u7684\u4e00\u7b14\u8d26\u3002\u4e2a\u4eba\u5f00\u53d1\u8005\u505a\u4e2a\u5c0f\u5de5\u5177\u3001\u8dd1\u4e2a Agent \u6d41\u7a0b\uff0c\u4e00\u5929\u51e0\u767e\u6b21\u8c03\u7528\u5f88\u6b63\u5e38\uff0c\u4e00\u4e2a\u6708\u4e0b\u6765\u96f6\u96f6\u788e\u788e\u4e5f\u662f\u4e00\u7b14\u94b1\u3002<\/p>\n\n\n\n

\u6240\u4ee5\u9009 API \u5e73\u53f0\uff0c\u9664\u4e86\u770b\u6a21\u578b\u591f\u4e0d\u591f\u5f3a\uff0c\u514d\u8d39\u989d\u5ea6\u591f\u4e0d\u591f\u7528\u4e5f\u5f88\u5173\u952e\u3002<\/p>\n\n\n\n

\u963f\u91cc\u7684\u9b54\u642d\u793e\u533a\uff08ModelScope\uff09<\/strong>\u5728\u8fd9\u65b9\u9762\u7ed9\u5f97\u76f8\u5f53\u5927\u65b9\u2014\u2014\u6bcf\u5929 2000 \u6b21\u514d\u8d39\u8c03\u7528\uff0c\u8986\u76d6 Qwen3\u3001DeepSeek-R1\u3001GLM-4.5 \u7b49\u4e3b\u6d41\u6a21\u578b\uff0c\u800c\u4e14\u652f\u6301 OpenAI \u517c\u5bb9\u534f\u8bae\uff0c\u63a5\u5165\u6210\u672c\u51e0\u4e4e\u4e3a\u96f6\u3002<\/p>\n\n\n\n

\u8fd9\u7bc7\u6587\u7ae0\u5c31\u628a\u9b54\u642d\u7684\u514d\u8d39\u989d\u5ea6\u89c4\u5219\u3001\u652f\u6301\u7684\u6a21\u578b\u3001\u8c03\u7528\u65b9\u5f0f\u3001\u9002\u914d\u5e73\u53f0\uff0c\u4e00\u6b21\u8bf4\u6e05\u695a\u3002<\/p>\n\n\n\n

\ud83e\udd16 \u9b54\u642d\u662f\u4ec0\u4e48\uff1f\u4e2d\u56fd\u7248 Hugging Face<\/h2>\n\n\n\n

\u9b54\u642d\u793e\u533a\uff08ModelScope\uff09<\/strong>\u662f\u963f\u91cc\u4e91\u65d7\u4e0b\u7684\u5f00\u6e90 AI \u6a21\u578b\u793e\u533a\uff0c\u5b9a\u4f4d\u662f”AI \u754c\u7684 GitHub”\u3002<\/p>\n\n\n\n

\u6838\u5fc3\u6570\u636e<\/th>\u8be6\u60c5<\/th><\/tr><\/thead>
\u6a21\u578b\u6570\u91cf<\/td>2 \u4e07+ \u4e2a\u6a21\u578b<\/td><\/tr>
\u514d\u8d39 API<\/td>\u6bcf\u5929\u603b\u8ba1 2000 \u6b21\u8c03\u7528<\/td><\/tr>
\u517c\u5bb9\u534f\u8bae<\/td>OpenAI + Anthropic<\/td><\/tr>
\u8986\u76d6\u9886\u57df<\/td>LLM \/ \u591a\u6a21\u6001 \/ \u6587\u751f\u56fe \/ \u8bed\u97f3 \/ \u4ee3\u7801<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

\ud83c\udfaf \u514d\u8d39\u989d\u5ea6\u5230\u5e95\u600e\u4e48\u7b97\uff1f<\/h2>\n\n\n\n
\u9879\u76ee<\/th>\u8be6\u60c5<\/th><\/tr><\/thead>
\u6bcf\u65e5\u603b\u989d\u5ea6<\/td>\u6240\u6709\u6a21\u578b\u5408\u8ba1 2000 \u6b21<\/td><\/tr>
\u5404\u6a21\u578b\u72ec\u7acb\u9650\u989d<\/td>\u6bcf\u4e2a\u6a21\u578b\u6709\u5404\u81ea\u7684\u6bcf\u65e5\u4e0a\u9650<\/td><\/tr>
\u91cd\u7f6e\u65f6\u95f4<\/td>\u6bcf\u5929 0 \u70b9<\/td><\/tr>
\u5b9e\u540d\u8981\u6c42<\/td>\u9700\u7ed1\u5b9a\u963f\u91cc\u4e91\u8d26\u53f7<\/td><\/tr>
\u8d85\u989d\u6263\u8d39\uff1f<\/td>\u4e0d\u4f1a\uff01\u8fd4\u56de 429 \u9519\u8bef<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

\u26a0\ufe0f \u91cd\u70b9<\/strong>\uff1a2000 \u6b21\u662f\u6240\u6709\u6a21\u578b\u52a0\u5728\u4e00\u8d77\u7684\u603b\u989d\u5ea6\u3002\u6bcf\u4e2a\u6a21\u578b\u6709\u5404\u81ea\u7684\u6bcf\u65e5\u4e0a\u9650\uff0c\u6bd4\u5982 DeepSeek-R1 \u6bcf\u5929\u7ea6 200 \u6b21\uff0c\u6587\u751f\u56fe\u6a21\u578b\u6bcf\u5929\u7ea6 50 \u6b21\u3002<\/p>\n\n\n\n

\u4e09\u6b65\u62ff\u5230\u514d\u8d39\u989d\u5ea6<\/h3>\n\n\n\n
    \n
  1. \u6ce8\u518c<\/strong>\uff1a\u8bbf\u95ee modelscope.cn<\/a><\/li>\n
  2. \u7ed1\u5b9a\u963f\u91cc\u4e91<\/strong>\uff1a\u5fc5\u987b\u7ed1\u5b9a<\/li>\n
  3. \u83b7\u53d6 Token<\/strong>\uff1a\u4e2a\u4eba\u4e2d\u5fc3 \u2192 \u8bbf\u95ee\u4ee4\u724c \u2192 \u521b\u5efa\u4ee4\u724c \u2192 \u52fe\u9009”\u5927\u6a21\u578b\u63a8\u7406”\u6743\u9650<\/li>\n<\/ol>\n\n\n\n

    \ud83e\udde0 \u652f\u6301\u54ea\u4e9b\u6a21\u578b\uff1f<\/h2>\n\n\n\n

    \u5927\u8bed\u8a00\u6a21\u578b<\/h3>\n\n\n\n
    \u6a21\u578b<\/th>\u8bf4\u660e<\/th>\u6bcf\u65e5\u514d\u8d39<\/th><\/tr><\/thead>
    Qwen3-Coder-480B<\/td>\u6700\u65b0\u4ee3\u7801\u6a21\u578b<\/td>~500 \u6b21<\/td><\/tr>
    Qwen3-235B-A22B<\/td>MoE \u65d7\u8230\u7248<\/td>~500 \u6b21<\/td><\/tr>
    DeepSeek-R1<\/td>\u6df1\u5ea6\u63a8\u7406\u6a21\u578b<\/td>~200 \u6b21<\/td><\/tr>
    DeepSeek-V3<\/td>\u6700\u65b0\u901a\u7528\u6a21\u578b<\/td>~500 \u6b21<\/td><\/tr>
    GLM-4.5<\/td>\u667a\u8c31 AI \u5927\u6a21\u578b<\/td>~500 \u6b21<\/td><\/tr>
    MiniMax-M1<\/td>\u957f\u6587\u672c\u6a21\u578b<\/td>~500 \u6b21<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u6587\u751f\u56fe\u6a21\u578b<\/h3>\n\n\n\n
    \u6a21\u578b<\/th>\u8bf4\u660e<\/th>\u6bcf\u65e5\u514d\u8d39<\/th><\/tr><\/thead>
    FLUX.1-dev<\/td>\u6700\u5f3a\u5f00\u6e90\u6587\u751f\u56fe<\/td>~50 \u6b21<\/td><\/tr>
    Stable Diffusion 3<\/td>\u7ecf\u5178\u6587\u751f\u56fe<\/td>~50 \u6b21<\/td><\/tr>
    \u5404\u79cd LoRA<\/td>1000+ \u98ce\u683c\u5316\u6a21\u578b<\/td>~50 \u6b21<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \u591a\u6a21\u6001 + \u8bed\u97f3<\/h3>\n\n\n\n
    \u7c7b\u578b<\/th>\u4ee3\u8868\u6a21\u578b<\/th><\/tr><\/thead>
    \u591a\u6a21\u6001\u7406\u89e3<\/td>Qwen2.5-VL-72B \/ Qwen-Omni<\/td><\/tr>
    \u8bed\u97f3\u8bc6\u522b<\/td>SenseVoice \/ FunASR<\/td><\/tr>
    \u8bed\u97f3\u5408\u6210<\/td>CosyVoice<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    \ud83d\udcbb \u600e\u4e48\u8c03\u7528\uff1f<\/h2>\n\n\n\n

    \u65b9\u5f0f\u4e00\uff1aOpenAI \u517c\u5bb9\u63a5\u53e3 \u2b50\u63a8\u8350<\/h3>\n\n\n\n
    from openai import OpenAI\nclient = OpenAI(\n    api_key=\"\u4f60\u7684\u9b54\u642dToken\",\n    base_url=\"https:\/\/api-inference.modelscope.cn\/v1\/\"\n)\nresponse = client.chat.completions.create(\n    model=\"Qwen\/Qwen3-235B-A22B-Instruct\",\n    messages=[{\"role\": \"user\", \"content\": \"\u7528Python\u5199\u4e00\u4e2a\u5feb\u901f\u6392\u5e8f\"}],\n    stream=True\n)<\/code><\/pre>\n\n\n\n
    \u65b9\u5f0f\u4e8c\uff1a\u6587\u751f\u56fe API<\/h3>\n\n\n\n
    import requests\nresponse = requests.post(\n    \"https:\/\/api-inference.modelscope.cn\/v1\/images\/generations\",\n    headers={\"Authorization\": \"Bearer \u4f60\u7684\u9b54\u642dToken\"},\n    json={\n        \"model\": \"MAILAND\/majicflus_v1\",\n        \"prompt\": \"\u4e00\u53ea\u7a7f\u897f\u88c5\u7684\u9f99\u867e\u5728\u5199\u4ee3\u7801\"\n    }\n)<\/code><\/pre>\n\n\n\n
    \ud83d\udd17 \u80fd\u7528\u5728\u54ea\u4e9b\u5e73\u53f0\uff1f<\/h2>\n\n\n\n
    \u5e73\u53f0\/\u5de5\u5177<\/th>\u63a5\u5165\u65b9\u5f0f<\/th>\u96be\u5ea6<\/th><\/tr><\/thead>
    Claude Code<\/td>\u8bbe\u7f6e\u73af\u5883\u53d8\u91cf<\/td>\u2b50 \u7b80\u5355<\/td><\/tr>
    OpenClaw<\/td>\u81ea\u5b9a\u4e49 API Provider<\/td>\u2b50 \u7b80\u5355<\/td><\/tr>
    Cursor<\/td>Settings \u2192 Custom API<\/td>\u2b50\u2b50 \u4e2d\u7b49<\/td><\/tr>
    Dify<\/td>\u6a21\u578b\u4f9b\u5e94\u5546 \u2192 OpenAI \u517c\u5bb9<\/td>\u2b50 \u7b80\u5355<\/td><\/tr>
    n8n<\/td>ModelScope \u63d2\u4ef6<\/td>\u2b50 \u7b80\u5355<\/td><\/tr>
    LangChain<\/td>ChatOpenAI(base_url=…)<\/td>\u2b50 \u7b80\u5355<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
    \ud83c\udfa8 \u9b54\u7c92\u7cfb\u7edf\uff08AIGC \u521b\u4f5c\uff09<\/h2>\n\n\n\n
    \u83b7\u53d6\u9014\u5f84<\/th>\u6bcf\u65e5\u989d\u5ea6<\/th><\/tr><\/thead>
    \u6bcf\u65e5\u767b\u5f55<\/td>150 \u9b54\u7c92<\/td><\/tr>
    \u521b\u4f5c\u8fd4\u56fe\u4efb\u52a1<\/td>+200 \u9b54\u7c92<\/td><\/tr>
    \u70b9\u8d5e\u8bc4\u8bba\u4efb\u52a1<\/td>+50 \u9b54\u7c92<\/td><\/tr>
    \u6a21\u578b\u4f7f\u7528\u5956\u52b1<\/td>+200 \u9b54\u7c92<\/td><\/tr>
    \u6bcf\u65e5\u6700\u9ad8<\/td>\u7ea6 600 \u9b54\u7c92<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
    \ud83d\udcbb \u514d\u8d39 GPU Notebook + \u6c38\u4e45\u514d\u8d39\u90e8\u7f72<\/h2>\n\n\n\n
    JupyterLab + WebIDE + Terminal\uff0c\u514d\u8d39 GPU \u7b97\u529b\u3002\u8fd8\u80fd\u628a\u6a21\u578b\u4e00\u952e\u90e8\u7f72\u6210\u5728\u7ebf\u670d\u52a1\uff0c\u6c38\u4e45\u514d\u8d39\u3002<\/p>\n\n\n\n
    \u26a0\ufe0f \u907f\u5751\u6307\u5357<\/h2>\n\n\n\n
    \u5751\u4e00\uff1a\u6ca1\u7ed1\u5b9a\u963f\u91cc\u4e91\u5c31\u8c03\u7528<\/strong> \u2192 \u53bb\u9b54\u642d”\u4e2a\u4eba\u8bbe\u7f6e”\u7ed1\u5b9a\u3002<\/p>\n\n\n\n
    \u5751\u4e8c\uff1a2000 \u6b21\u989d\u5ea6\u4e0d\u591f\u7528\uff1f<\/strong> \u2192 \u591a\u5e73\u53f0\u7ec4\u5408\uff1a\u9b54\u642d 2000 \u6b21 + \u706b\u5c71 200 \u4e07 Token + \u767e\u70bc 100 \u4e07 Token\u3002<\/p>\n\n\n\n
    \u5751\u4e09\uff1a\u6df1\u5ea6\u63a8\u7406\u6a21\u578b\u989d\u5ea6\u5c11<\/strong> \u2192 \u7b80\u5355\u4efb\u52a1\u7528 Qwen3 \u6216 DeepSeek-V3\u3002<\/p>\n\n\n\n
    \u5751\u56db\uff1a\u8d85\u989d\u4f1a\u81ea\u52a8\u6263\u8d39\u5417\uff1f<\/strong> \u2192 \u4e0d\u4f1a\uff01\u8d85\u989d\u8fd4\u56de 429 \u9519\u8bef\u3002<\/p>\n\n\n\n
    \ud83d\udcdd \u4e00\u53e5\u8bdd\u603b\u7ed3<\/h2>\n\n\n\n
    \u9b54\u642d\u793e\u533a = \u4e2d\u56fd\u7248 Hugging Face<\/p>\n\n\n\n