{"id":67169,"date":"2026-01-28T13:06:51","date_gmt":"2026-01-28T05:06:51","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/67169.html"},"modified":"2026-01-28T13:06:51","modified_gmt":"2026-01-28T05:06:51","slug":"%e5%bc%80%e6%ba%90ai%e6%a8%a1%e5%9e%8b%e7%bb%88%e6%9e%81%e5%af%bc%e8%88%aa%ef%bc%9a%e4%bb%8ellm%e5%88%b0%e5%a4%9a%e6%a8%a1%e6%80%81%ef%bc%8c%e4%b8%80%e6%96%87%e6%95%99%e4%bd%a0%e5%a6%82%e4%bd%95","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/67169.html","title":{"rendered":"\u5f00\u6e90AI\u6a21\u578b\u7ec8\u6781\u5bfc\u822a\uff1a\u4eceLLM\u5230\u591a\u6a21\u6001\uff0c\u4e00\u6587\u6559\u4f60\u5982\u4f55\u9009\u578b\u4e0e\u843d\u5730"},"content":{"rendered":"

\u4e00\u3001\u5e38\u7528\u5f00\u6e90\u5927\u8bed\u8a00\u6a21\u578b\u5bf9\u6bd4（LLM）<\/span><\/h3>\n

1. \u4e3b\u6d41\u901a\u7528 LLM：\u600e\u4e48\u9009？<\/span><\/h4>\n

\u573a\u666f：\u5bf9\u8bdd\u3001\u5199\u4f5c\u3001\u95ee\u7b54\u3001\u5206\u6790\u3001RAG\u3001\u5927\u90e8\u5206\u901a\u7528\u5e94\u7528\u3002<\/span><\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n
\n

\u6a21\u578b\u5bb6\u65cf<\/span><\/p>\n<\/td>\n

\n

\u4ee3\u8868\u578b\u53f7（\u5f00\u6e90\u6743\u91cd）<\/span><\/p>\n<\/td>\n

\n

\u7efc\u5408\u5b9e\u529b & \u7279\u70b9<\/span><\/p>\n<\/td>\n

\n

\u9002\u5408\u4f60\u5728\u4ec0\u4e48\u573a\u666f\u7528<\/span><\/p>\n<\/td>\n<\/tr>\n

\n

Qwen（\u901a\u4e49\u5343\u95ee，\u963f\u91cc）\u200b<\/span><\/p>\n<\/td>\n

\n

Qwen2 \/ Qwen2.5 \/ <\/span>Qwen3-235B-A22B<\/span>\u00a0\u7b49<\/span><\/p>\n<\/td>\n

\n

\u5728\u65af\u5766\u798f HELM \/ MMLU \u7b49\u6743\u5a01\u699c\u5355\u4e0a，Qwen2 \/ Qwen3 \u7cfb\u5217\u662f\u76ee\u524d\u6700\u5f3a\u7684\u5f00\u6e90\u6a21\u578b\u4e4b\u4e00，\u591a\u6b21\u538b\u8fc7 Llama \u7cfb\u5217；\u5728 Hugging Face \u5f00\u6e90\u699c\u5355\u4e2d，\u57fa\u4e8e Qwen \u7684\u884d\u751f\u6a21\u578b\u6570\u91cf\u548c\u4e0b\u8f7d\u91cf\u957f\u671f\u7b2c\u4e00，[HF \u5b98\u65b9\u7edf\u8ba1\u79f0 Qwen \u884d\u751f\u6a21\u578b\u5df2 20 \u4e07+\u3001\u4e0b\u8f7d\u91cf 10 \u4ebf+，\u7a33\u5c45\u5168\u7403\u5f00\u6e90 TOP<\/span>1<\/span>\u3002\u4e2d\u6587\u80fd\u529b\u5c24\u5176\u5f3a，\u82f1\u6587\u548c\u591a\u8bed\u8a00\u4e5f\u5f88\u4eae\u773c\u3002<\/span><\/p>\n<\/td>\n

\n

\u4e2d\u6587\u4e3a\u4e3b + \u6709\u4e00\u5b9a\u82f1\u6587\/\u591a\u8bed\u9700\u6c42<\/span>；\u516c\u53f8\/\u56e2\u961f\u60f3\u8981\u8ddf\u7740\u201c\u5f00\u6e90\u4e3b\u6d41\u751f\u6001\u201d\u8d70，\u4f18\u5148\u8003\u8651 Qwen \u7cfb\u5217\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

\n

Llama（Meta）\u200b<\/span><\/p>\n<\/td>\n

\n

Llama 3 \/ Llama 4 Scout \/ Maverick \/ Behemoth<\/span><\/p>\n<\/td>\n

\n

Llama \u4e00\u76f4\u662f\u5f00\u6e90 LLM \u7684\u201c\u9ed8\u8ba4\u57fa\u51c6\u201d\u3002Llama 3 \u65f6\u4ee3\u540c\u4f53\u91cf\u8868\u73b0\u6781\u5f3a；Llama 4 \u53c8\u5728\u591a\u6a21\u6001\u548c\u957f\u4e0a\u4e0b\u6587\u4e0a\u5927\u5e45\u5347\u7ea7，\u5b98\u65b9\u5ba3\u79f0 Llama 4 Behemoth \u5728\u591a\u4e2a STEM（\u7406\u5de5）\u57fa\u51c6\u4e0a\u8d85 GPT4.5 \u7b49\u95ed\u6e90\u6a21\u578b，\u591a\u6a21\u6001\u3001\u957f\u4e0a\u4e0b\u6587（\u6700\u9ad8\u5230 10M tokens）\u4e5f\u975e\u5e38\u62d4\u5c16\u3002<\/span><\/p>\n<\/td>\n

\n

\u9700\u8981\u56fd\u9645\u751f\u6001\u3001\u517c\u5bb9\u6027\u548c\u6587\u6863<\/span>（\u82f1\u6587\u793e\u533a\u3001\u5de5\u5177\u6700\u591a），\u6216\u8981\u505a<\/span>\u591a\u6a21\u6001（\u56fe\u6587）+ \u957f\u6587\u6863<\/span>，Llama 4 \u662f\u76ee\u524d\u6700\u4fdd\u9669\u7684\u9009\u62e9\u4e4b\u4e00\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

\n

DeepSeek（\u6df1\u5ea6\u6c42\u7d22）\u200b<\/span><\/p>\n<\/td>\n

\n

DeepSeekV3 \/ V3.2 \/ R1 \/ V4（\u90e8\u5206\u5f00\u6e90\u6743\u91cd）\u200b<\/span><\/p>\n<\/td>\n

\n

DeepSeek \u7cfb\u5217\u5728\u6570\u5b66\u63a8\u7406\u3001\u4ee3\u7801\u548c\u590d\u6742\u63a8\u7406\u4efb\u52a1\u4e0a\u975e\u5e38\u6fc0\u8fdb\u3002\u516c\u5f00\u6570\u636e\u8868\u660e：DeepSeekV3.2 \u5728 MATH500\u3001HumanEval \u7b49\u96be\u5ea6\u9ad8\u7684\u6570\u5b66\u548c\u4ee3\u7801\u57fa\u51c6\u4e0a\u8fbe\u5230\u91d1\u724c\u6c34\u5e73，\u591a\u6b21\u88ab\u82f1\u6587\u5a92\u4f53\u79f0\u4e3a\u201c\u5728\u5f00\u6e90\u7aef\u6253\u7834\u95ed\u6e90\u5784\u65ad\u201d\u7684\u4ee3\u8868\u3002R1 \/ V4 \u66f4\u504f\u201c\u601d\u7ef4\u94fe\u5f0f\u63a8\u7406\u201d（reasoning models）\u3002<\/span><\/p>\n<\/td>\n

\n

\u9700\u8981\u6781\u5f3a\u63a8\u7406\u548c\u4ee3\u7801\u80fd\u529b<\/span>（\u4f8b\u5982\u81ea\u52a8\u5237\u9898\u3001\u79d1\u7814\u63a8\u5bfc\u3001\u590d\u6742\u4ee3\u7801\u91cd\u6784），DeepSeek \u7cfb\u5217\u4f18\u5148\u8003\u8651\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

\n

Mixtral（Mistral AI）\u200b<\/span><\/p>\n<\/td>\n

\n

Mixtral 8\u00d77B \/ 8\u00d722B<\/span><\/p>\n<\/td>\n

\n

\u7a00\u758f MoE \u67b6\u6784，\u6fc0\u6d3b\u90e8\u5206\u4e13\u5bb6，\u63a8\u7406\u901f\u5ea6\u5feb\u3001\u541e\u5410\u9ad8\u3002\u5b98\u65b9\u6570\u636e\u8868\u660e Mixtral 8\u00d722B \u5728 MMLU\u3001GSM8K（\u6570\u5b66）\u3001HumanEval（\u4ee3\u7801）\u4e0a\u5168\u9762\u8d85\u8fc7 Llama 2 70B \u7b49\u8001\u4e00\u4ee3\u6a21\u578b，\u540c\u65f6\u5728\u63a8\u7406\u901f\u5ea6\/\u6210\u672c\u4e0a\u6709\u660e\u663e\u4f18\u52bf\u3002<\/span><\/p>\n<\/td>\n

\n

\u5e0c\u671b\u517c\u987e\u6027\u80fd\u548c\u6210\u672c<\/span>，\u5728\u4e00\u5230\u4e24\u5f20\u9ad8\u7aef\u5361\u4e0a\u8dd1\u4e00\u4e2a\u201c\u5f88\u80fd\u6253\u201d\u7684\u901a\u7528\u6a21\u578b，Mixtral \u503c\u5f97\u4f18\u5148\u8bd5\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

\n

Gemma（Google）\u200b<\/span><\/p>\n<\/td>\n

\n

Gemma 2\/3（1B\u201327B \u7cfb\u5217）<\/span><\/p>\n<\/td>\n

\n

\u65b0\u4e00\u4ee3\u8f7b\u91cf\u5f00\u6e90\u6a21\u578b，\u591a\u4e2a\u57fa\u51c6\u4e0a\u5df2\u63a5\u8fd1\u751a\u81f3\u8d85\u8fc7\u90e8\u5206\u81ea\u5bb6 Gemini Flash；\u4e0a\u4e0b\u6587\u6700\u957f\u5230 128K tokens，\u8c37\u6b4c\u4e13\u95e8\u9488\u5bf9\u5b89\u5168\u548c\u6613\u90e8\u7f72\u505a\u4e86\u4f18\u5316\u3002<\/span><\/p>\n<\/td>\n

\n

\u5bf9\u8c37\u6b4c\u751f\u6001\u53cb\u597d<\/span>（GCP\u3001Vertex AI）\u6216\u9700\u8981\u6bd4\u8f83\u5c0f\u4f46\u9760\u8c31\u7684\u6a21\u578b（1B\u20134B \u7ea7），Gemma \u662f\u4e00\u4e2a\u5e72\u51c0\u3001\u5de5\u7a0b\u5316\u8f83\u597d\u7684\u9009\u62e9\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

\n

Phi3（\u5fae\u8f6f）\u200b<\/span><\/p>\n<\/td>\n

\n

Phi3 Mini \/ Small \/ Medium<\/span><\/p>\n<\/td>\n

\n

\u5fae\u8f6f\u5b98\u65b9\u7ed9\u5b9a\u4f4d\u662f\u201c\u5c0f\u800c\u5f3a\u201d：\u6700\u5c0f\u7684 Phi3Mini \u4ec5 3.8B \u53c2\u6570，\u5728\u5f88\u591a\u57fa\u51c6\u4e0a\u903c\u8fd1\u751a\u81f3\u5bf9\u9f50 Mixtral 8\u00d77B\u3001GPT3.5 \u7b49\u5927\u6a21\u578b，MMLU \u5927\u7ea6\u5728 69% \u5de6\u53f3\u3002\u7279\u522b\u9002\u5408\u8fb9\u7f18\u8bbe\u5907\u3001\u672c\u5730\u90e8\u7f72\u3002<\/span><\/p>\n<\/td>\n

\n

\u672c\u5730\u7535\u8111 \/ \u624b\u673a \/ \u5d4c\u5165\u5f0f\u8bbe\u5907<\/span>\u00a0\u60f3\u8dd1\u4e00\u4e2a\u201c\u771f\u80fd\u7528\u201d\u7684\u5c0f\u6a21\u578b，Phi3 \u662f\u76ee\u524d\u5f00\u6e90\u91cc\u6700\u5f3a\u7684\u5c0f\u4f53\u91cf\u4e4b\u4e00\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

\n

ChatGLM \/ GLM（\u667a\u8c31 AI）\u200b<\/span><\/p>\n<\/td>\n

\n

ChatGLM6B \/ GLM4.x<\/span><\/p>\n<\/td>\n

\n

\u6700\u65e9\u4e00\u6279\u53ef\u5546\u7528\u7684\u4e2d\u6587\u5f00\u6e90\u5bf9\u8bdd\u6a21\u578b\u4e4b\u4e00；GLM4 \u7cfb\u5217\u7ee7\u7eed\u5f3a\u5316\u591a\u8bed\u79cd\u548c\u591a\u6a21\u6001，\u5728 32K\u2013128K \u4e0a\u4e0b\u6587\u3001\u56fe\u6587\u6df7\u5408\u3001\u56fd\u4ea7\u786c\u4ef6\u9002\u914d（\u534e\u4e3a\u6607\u817e\u7b49）\u65b9\u9762\u5e03\u5c40\u5f88\u91cd\u3002<\/span><\/p>\n<\/td>\n

\n

\u56fd\u5185\u653f\u4f01\u3001\u6559\u80b2\u3001\u91d1\u878d\u7b49\u573a\u666f<\/span>，\u9700\u8981\u56fd\u4ea7\u5316\u7a0b\u5ea6\u9ad8\u3001\u9002\u914d\u56fd\u4ea7\u82af\u7247\/\u79c1\u6709\u4e91，GLM\/ChatGLM \u503c\u5f97\u5355\u72ec\u5217\u4e00\u6863\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

\n

Yi \u7cfb\u5217（\u96f6\u4e00\u4e07\u7269）\u200b<\/span><\/p>\n<\/td>\n

\n

Yi34B \/ Yi1.534B<\/span><\/p>\n<\/td>\n

\n

Yi34B \u5728 MMLU \u4e0a\u5927\u7ea6 76.3%，\u8d85 Llama2 70B \/ Falcon180B \u7b49\u65e9\u671f\u5927\u6a21\u578b，\u5c24\u5176\u5728\u4e2d\u6587\u548c\u4ee3\u7801\u4efb\u52a1\u4e0a\u53e3\u7891\u4e0d\u9519，\u591a\u6b3e\u884d\u751f\u5fae\u8c03\u6a21\u578b\u4e5f\u5f88\u6d41\u884c\u3002<\/span><\/p>\n<\/td>\n

\n

\u60f3\u5728\u4e2d\u6587+\u82f1\u6587\u53cc\u8bed\u4e0a\u4fdd\u6301\u5e73\u8861<\/span>，\u53c8\u5e0c\u671b\u6a21\u578b\u672c\u8eab\u6bd4\u8f83\u201c\u5e72\u51c0\u201d，Yi34B \/ Yi1.534B \u662f\u4e00\u6761\u6210\u719f\u8def\u7ebf\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

\n

CommandR+（Cohere）\u200b<\/span><\/p>\n<\/td>\n

\n

CommandR+<\/span><\/p>\n<\/td>\n

\n

\u66f4\u504f\u4f01\u4e1a\u68c0\u7d22\u589e\u5f3a（RAG）\u573a\u666f，MMLU \u5728 75% \u5de6\u53f3，\u4f46\u5728 HellaSwag\u3001Winogrande \u7b49\u5e38\u8bc6\u63a8\u7406\u57fa\u51c6\u4e0a\u8868\u73b0\u5747\u8861，\u4e0a\u4e0b\u6587\u652f\u6301 128K tokens，\u975e\u5e38\u9002\u5408\u957f\u6587\u6863\u3001\u591a\u8f6e\u68c0\u7d22\u95ee\u7b54\u3002<\/span><\/p>\n<\/td>\n

\n

\u4f01\u4e1a\u77e5\u8bc6\u5e93\u3001\u6587\u6863\u95ee\u7b54\u3001RAG \u7cfb\u7edf<\/span>，\u5c24\u5176\u662f\u591a\u8bed\u8a00\u6587\u6863\u5e93，CommandR+ \u662f\u4e13\u95e8\u4e3a\u8fd9\u7c7b\u4efb\u52a1\u8bbe\u8ba1\u7684\u5f00\u6e90\/\u5f00\u6743\u6a21\u5f0f\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

\u7b80\u5316\u5efa\u8bae：<\/span><\/p>\n

    \n
  • \u505a\u4e2d\u6587\u4ea7\u54c1 \/ \u4e2d\u6587\u4e3a\u4e3b<\/span>：Qwen3 > DeepSeek > ChatGLM \/ GLM > Yi<\/span><\/li>\n
  • \u9700\u8981\u56fd\u9645\u5316\u3001\u591a\u6a21\u6001\u3001\u957f\u4e0a\u4e0b\u6587<\/span>：Llama 4 \/ Qwen3 \/ DeepSeek<\/span><\/li>\n
  • \u5355\u673a\/\u672c\u5730\u60f3\u8dd1\u5f97\u52a8<\/span>：Phi3 \/ Gemma 3 \u5c0f\u578b\u53f7 \/ Qwen1.5B\/4B \u7b49<\/span><\/li>\n
  • \u4f01\u4e1a\u77e5\u8bc6\u5e93 + RAG<\/span>：CommandR+ \/ Llama 4 \/ Qwen3<\/span><\/li>\n<\/ul>\n

    \u4e8c\u3001\u5e38\u7528\u5f00\u6e90\u591a\u6a21\u6001 & \u56fe\u50cf\u751f\u6210\u6a21\u578b<\/span><\/h3>\n

    1. \u6587\u672c\u2192\u56fe\u50cf\u6a21\u578b<\/span><\/h4>\n\n\n\n\n\n\n\n
    \n

    \u6a21\u578b<\/span><\/p>\n<\/td>\n

    		\n

    \u7279\u70b9<\/span><\/p>\n<\/td>\n

    		\n

    \u9002\u7528\u573a\u666f<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    Stable Diffusion 3.x \/ 3.5<\/span><\/p>\n<\/td>\n

    		\n

    \u7ecf\u5178\u5f00\u6e90\u751f\u56fe\u7cfb，3.5 \u5728\u753b\u8d28\u548c\u6307\u4ee4\u7406\u89e3\u4e0a\u6bd4 2.x \u5927\u5e45\u63d0\u5347，\u793e\u533a\u63d2\u4ef6\u3001UI（\u5982 Automatic1111\u3001ComfyUI）\u6700\u4e30\u5bcc，\u5f88\u591a\u7b2c\u4e09\u65b9 benchmark \u7ed9\u51fa\u6781\u9ad8\u6027\u4ef7\u6bd4\u8bc4\u5206\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u901a\u7528\u56fe\u7247\u751f\u6210\u3001\u63d2\u753b\u3001\u6d77\u62a5\u3001\u6e38\u620f\u539f\u753b；\u9700\u8981\u5b8c\u5168\u81ea\u6258\u7ba1\u3001\u672c\u5730\u5316\u7684\u9996\u9009\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    FLUX.1 \/ FLUX.2 \u7cfb\u5217<\/span><\/p>\n<\/td>\n

    		\n

    2025\u20132026 \u5e74\u6700\u53d7\u5173\u6ce8\u7684\u5f00\u6e90\u751f\u56fe\u65cf\u4e4b\u4e00，\u591a\u7bc7\u8bc4\u6d4b\u8ba4\u4e3a FLUX.2 Pro \u753b\u8d28\u548c Prompt \u5fe0\u5b9e\u5ea6\u5df2\u7ecf\u8ffd\u5e73\u751a\u81f3\u8d85\u8fc7\u5f88\u591a\u95ed\u6e90\u5546\u7528\u6a21\u578b[20][21]\u3002Kontext \u7248\u672c\u9488\u5bf9\u591a\u56fe\u62fc\u63a5\u3001\u80cc\u666f\u4e00\u81f4\u5316\u7b49\u505a\u4e86\u4f18\u5316\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u5bf9\u753b\u8d28\u3001\u6587\u5b57\u51c6\u786e\u5ea6\u548c\u53ef\u63a7\u6027\u8981\u6c42\u9ad8\u7684\u751f\u4ea7\u7ea7\u573a\u666f，\u6bd4\u5982\u5e7f\u544a\u3001\u7535\u5546\u8be6\u60c5\u9875\u3001\u5f71\u89c6\u5206\u955c\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    GLMImage（\u667a\u8c31 + \u534e\u4e3a）\u200b<\/span><\/p>\n<\/td>\n

    		\n

    \u5168\u6d41\u7a0b\u5728\u56fd\u4ea7\u6607\u817e\u82af\u7247\u4e0a\u8bad\u7ec3\u7684\u591a\u6a21\u6001\u56fe\u50cf\u751f\u6210\u6a21\u578b，\u5f00\u6e90\u4e0d\u5230 24 \u5c0f\u65f6\u767b\u9876 Hugging Face trending \u699c；\u5728\u590d\u6742\u4e2d\u6587\u6587\u5b57\u6e32\u67d3\u3001LOGO\u3001UI \u5e03\u5c40\u7b49\u65b9\u9762\u8868\u73b0\u7a81\u51fa\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u4e2d\u6587\u6587\u5b57\u56fe\u7247\u3001\u6d77\u62a5\u3001PPT \u5c01\u9762\u3001\u6559\u80b2\u8bfe\u4ef6<\/span>\u00a0\u7b49\u9700\u8981\u4e2d\u6587\u5b57\u4f53\u6e05\u6670\u7684\u573a\u666f\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    Hunyuan Image 3.0（\u817e\u8baf\u6df7\u5143）\u200b<\/span><\/p>\n<\/td>\n

    		\n

    80B \u7ea7 MoE \u67b6\u6784，\u9488\u5bf9\u8d85\u9ad8\u5206\u8fa8\u7387\u548c\u591a\u98ce\u683c\u505a\u4e86\u4f18\u5316，\u591a\u5bb6\u8bc4\u6d4b\u8ba4\u4e3a\u5728\u5f00\u6e90\u6a21\u578b\u4e2d\u5c5e\u4e8e\u201c\u5927\u800c\u5168\u201d\u7684\u65d7\u8230，\u591a\u6a21\u6001\u7406\u89e3\u80fd\u529b\u548c\u56fe\u50cf\u7ec6\u8282\u90fd\u5f88\u5f3a\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u9ad8\u7aef\u5546\u4e1a\u63d2\u753b\u3001\u5e7f\u544a\u3001\u5370\u5237\u7ea7\u56fe\u7247，\u6216\u8005\u9700\u8981\u591a\u98ce\u683c\u521b\u610f\u8f93\u51fa\u7684\u5927\u516c\u53f8\u56e2\u961f\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    \n

    \n

    \u4e09\u3001\u5e38\u7528\u5f00\u6e90\u8bed\u97f3\u6a21\u578b（TTS\/ASR）<\/span><\/h3>\n\n\n\n\n\n\n
    \n

    \u6a21\u578b<\/span><\/p>\n<\/td>\n

    		\n

    \u7c7b\u578b<\/span><\/p>\n<\/td>\n

    		\n

    \u4eae\u70b9<\/span><\/p>\n<\/td>\n

    		\n

    \u9002\u7528\u573a\u666f<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    Fish Speech V1.5 \/ FishAudioS1<\/span><\/p>\n<\/td>\n

    		\n

    \u6587\u672c\u8f6c\u8bed\u97f3 TTS<\/span><\/p>\n<\/td>\n

    		\n

    \u91c7\u7528 DualAR \u67b6\u6784 + \u53cc\u81ea\u56de\u5f52 Transformer，\u652f\u6301\u4e2d\/\u82f1\/\u65e5\u7b49\u591a\u8bed\u79cd，\u72ec\u7acb\u8bc4\u6d4b\u4e2d ELO \u6392\u540d\u9760\u524d，\u4e2d\u6587\u5b57\u7b26 CER \u7ea6 1.3%；\u60c5\u611f\u3001\u8bed\u8c03\u63a7\u5236\u7ec6\u817b\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u9700\u8981\u9ad8\u8d28\u91cf\u3001\u591a\u60c5\u7eea\u3001\u591a\u8bed\u79cd\u914d\u97f3\u7684\u5e94\u7528：\u6709\u58f0\u4e66\u3001\u77ed\u89c6\u9891\u89e3\u8bf4\u3001\u865a\u62df\u4e3b\u64ad\u7b49\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    OpenAudioS1<\/span><\/p>\n<\/td>\n

    		\n

    \u60c5\u611f TTS<\/span><\/p>\n<\/td>\n

    		\n

    \u5f3a\u8c03\u60c5\u7eea\u8868\u8fbe（\u54ed\u3001\u7b11\u7b49）\u3001\u5b9e\u65f6\u6027\u548c\u5f00\u6e90\u53ef\u5b9a\u5236，\u662f\u793e\u533a\u5e38\u88ab\u62ff\u6765\u548c Fish Speech \u5bf9\u6bd4\u7684\u4e00\u7ebf\u6a21\u578b\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u5b9e\u65f6\u8bed\u97f3\u4ee3\u7406\u3001\u966a\u4f34\u578b\u673a\u5668\u4eba\u3001\u6e38\u620f NPC\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    Whisper \/ Whisper v3<\/span><\/p>\n<\/td>\n

    		\n

    \u8bed\u97f3\u8bc6\u522b ASR<\/span><\/p>\n<\/td>\n

    		\n

    OpenAI \u51fa\u7684\u901a\u7528\u8bed\u97f3\u8bc6\u522b\/\u7ffb\u8bd1\u6a21\u578b，\u5df2\u88ab\u65e0\u6570\u5f00\u6e90\u9879\u76ee\u96c6\u6210；\u591a\u8bed\u8a00\u8bc6\u522b\/\u7ffb\u8bd1\u80fd\u529b\u5f3a，\u5f00\u6e90\u534f\u8bae\u5bbd\u677e\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u4f1a\u8bae\u8bb0\u5f55\u3001\u5b57\u5e55\u751f\u6210\u3001\u591a\u8bed\u8a00\u7ffb\u8bd1\u3001\u8bed\u97f3\u641c\u7d22\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    \n

    \n

    \u56db\u3001\u5e38\u7528\u5f00\u6e90\u6a21\u578b\u7f51\u7ad9 \/ \u5e73\u53f0<\/span><\/h3>\n

    1. \u56fd\u9645\u901a\u7528\u5e73\u53f0<\/span><\/h4>\n\n\n\n\n\n\n\n
    \n

    \u5e73\u53f0<\/span><\/p>\n<\/td>\n

    		\n

    \u4e3b\u8981\u7528\u9014<\/span><\/p>\n<\/td>\n

    		\n

    \u9002\u5408\u4eba\u7fa4<\/span><\/p>\n<\/td>\n

    		\n

    \u7f51\u5740<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    Hugging Face<\/span><\/p>\n<\/td>\n

    		\n

    \u6700\u5927\u7684\u5f00\u6e90\u6a21\u578b\u548c\u6570\u636e\u96c6\u5e73\u53f0，\u6258\u7ba1 50 \u4e07+ \u6a21\u578b（LLM\u3001CV\u3001\u97f3\u9891\u3001\u5f3a\u5316\u5b66\u4e60\u7b49\u5168\u54c1\u7c7b），\u63d0\u4f9b\u63a8\u7406 API\u3001Spaces Demo \u548c\u8bc4\u6d4b\u699c\u5355（\u5982 Open LLM Leaderboard）\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u6240\u6709\u505a AI \u7684\u4eba：\u4ece\u5b66\u751f\u5230\u4f01\u4e1a\u3002\u627e\u4efb\u4f55\u6a21\u578b，\u5148\u641c HF \u57fa\u672c\u662f\u5e38\u8bc6\u3002<\/span><\/p>\n<\/td>\n

    		\n

    https:\/\/huggingface.co<\/p>\n<\/td>\n<\/tr>\n

    \n

    GitHub<\/span><\/p>\n<\/td>\n

    		\n

    \u4ee3\u7801\u4e0e\u6a21\u578b\u6743\u91cd\u4ed3\u5e93，\u5f88\u591a\u56e2\u961f\u5148\u5728 GitHub \u53d1\u4ee3\u7801\u518d\u540c\u6b65\u5230 HF\/ModelScope\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u9700\u8981\u770b\u6e90\u7801\u3001\u8bad\u7ec3\u811a\u672c\u3001Issue\u3001\u793e\u533a\u8ba8\u8bba\u7684\u5f00\u53d1\u8005\u3002<\/span><\/p>\n<\/td>\n

    		\n

    https:\/\/github.com<\/p>\n<\/td>\n<\/tr>\n

    \n

    Kaggle Models \/ Datasets<\/span><\/p>\n<\/td>\n

    		\n

    \u4ee5\u7ade\u8d5b\u548c\u6570\u636e\u96c6\u8457\u79f0，\u73b0\u5728\u4e5f\u6258\u7ba1\u4e0d\u5c11\u6a21\u578b（\u5982 DeepSeekV3\u3001Phi3 \u7b49），\u4fbf\u4e8e\u76f4\u63a5\u5728 Notebook \u73af\u5883\u590d\u73b0\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u60f3\u5728 Notebook \u91cc\u5feb\u901f\u8bd5\u6a21\u578b\u7684\u6570\u636e\u79d1\u5b66\/\u5de5\u7a0b\u540c\u5b66\u3002<\/span><\/p>\n<\/td>\n

    		\n

    https:\/\/www.kaggle.com\/models<\/p>\n<\/td>\n<\/tr>\n

    \n

    OpenRouter \/ Together \/ Replicate \u7b49<\/span><\/p>\n<\/td>\n

    		\n

    \u805a\u5408\u591a\u4f9b\u5e94\u5546 LLM \u548c\u591a\u6a21\u6001\u6a21\u578b，\u901a\u8fc7\u7edf\u4e00 API \u8c03\u7528；\u4f60\u53ef\u4ee5\u7528\u4e00\u4e2a Key \u8c03 DeepSeek\u3001Qwen\u3001Llama\u3001Gemma \u7b49\u5404\u79cd\u6a21\u578b\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u4e0d\u60f3\u81ea\u5df1\u90e8\u7f72，\u53ea\u60f3\u4ee5\u4f4e\u6210\u672c\u7528\u201c\u5f00\u6e90\u6a21\u578b\u7684\u4e91\u6258\u7ba1\u7248\u672c\u201d\u7684\u5f00\u53d1\u8005\u3002<\/span><\/p>\n<\/td>\n

    		\n

    https:\/\/openrouter.ai\u00a0\/ <\/span>https:\/\/www.together.ai\u00a0\/ <\/span>https:\/\/replicate.com<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    2. \u4e2d\u6587\u751f\u6001\u4e3b\u5e73\u53f0<\/span><\/h4>\n\n\n\n\n\n\n\n
    \n

    \u5e73\u53f0<\/span><\/p>\n<\/td>\n

    		\n

    \u80cc\u666f & \u4eae\u70b9<\/span><\/p>\n<\/td>\n

    		\n

    \u9002\u7528\u573a\u666f<\/span><\/p>\n<\/td>\n

    		\n

    \u7f51\u5740<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    ModelScope（\u9b54\u642d\u793e\u533a，\u963f\u91cc）\u200b<\/span><\/p>\n<\/td>\n

    		\n

    \u9762\u5411\u4e2d\u6587\u5f00\u53d1\u8005\u7684\u6a21\u578b\u5e73\u53f0，\u652f\u6301\u6587\u672c\u3001\u56fe\u50cf\u3001\u89c6\u9891\u3001\u8bed\u97f3\u7b49\u591a\u6a21\u6001；\u4e0e Qwen \u7cfb\u5217\u6df1\u5ea6\u7ed1\u5b9a，\u5f88\u591a\u963f\u91cc\/\u56fd\u5185\u9ad8\u6821\/\u5382\u5546\u6a21\u578b\u9996\u53d1\u5728\u8fd9\u4e0a\u9762[28][29]\u3002\u63d0\u4f9b\u5728\u7ebf\u4f53\u9a8c\u3001Notebook\u3001\u8bad\u7ec3\/\u90e8\u7f72\u4e00\u4f53\u5de5\u5177\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u4e3b\u8981\u505a\u4e2d\u6587\u573a\u666f<\/span>，\u4e14\u5e0c\u671b\u4e0b\u8f7d\u901f\u5ea6\u3001\u6587\u6863\u3001\u793a\u4f8b\u4ee3\u7801\u90fd\u5bf9\u4e2d\u6587\u53cb\u597d，\u53ef\u4ee5\u4f18\u5148\u770b\u9b54\u642d\u3002<\/span><\/p>\n<\/td>\n

    		\n

    https:\/\/modelscope.cn<\/p>\n<\/td>\n<\/tr>\n

    \n

    ModelScope AI \/ \u963f\u91cc\u4e91 ModelStudio<\/span><\/p>\n<\/td>\n

    		\n

    \u9488\u5bf9\u4f01\u4e1a\u7684 MaaS \u5e73\u53f0，\u652f\u6301\u76f4\u63a5\u6258\u7ba1\u548c\u8c03\u7528 Qwen\u3001GLM\u3001Stable Diffusion \u7b49\u6a21\u578b，\u96c6\u6210\u963f\u91cc\u4e91\u5b89\u5168\u5ba1\u8ba1\u3001\u8ba1\u8d39\u548c\u8fd0\u7ef4\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u4e2d\u5927\u578b\u4f01\u4e1a\u60f3\u7528\u201c\u5f00\u653e\u6a21\u578b + \u4e91\u4e0a\u6258\u7ba1 + \u5408\u89c4\u5b89\u5168\u201d\u7684\u4e00\u4f53\u89e3\u51b3\u65b9\u6848\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u901a\u8fc7\u963f\u91cc\u4e91\u63a7\u5236\u53f0\u8fdb\u5165<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    \u767e\u5ea6\u98de\u6868 PaddlePaddle & AI Studio<\/span><\/p>\n<\/td>\n

    		\n

    \u4ee5\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6+\u4ea7\u4e1a\u7ea7\u6a21\u578b\u4e3a\u4e3b，\u878d\u5408 OCR\u3001NLP\u3001CV\u3001\u591a\u6a21\u6001\u7b49；\u5bf9\u4e8e\u653f\u5e9c\u3001\u91d1\u878d\u7b49\u56fd\u4ea7\u5316\u8981\u6c42\u9ad8\u7684\u573a\u666f，\u98de\u6868\u751f\u6001\u662f\u91cd\u8981\u9009\u9879\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u653f\u4f01\/\u7814\u7a76\u673a\u6784，\u7279\u522b\u662f\u6709\u56fd\u4ea7\u8f6f\u786c\u4ef6\u9002\u914d\u8981\u6c42\u7684\u9879\u76ee\u3002<\/span><\/p>\n<\/td>\n

    		\n

    https:\/\/www.paddlepaddle.org.cn<\/p>\n<\/td>\n<\/tr>\n

    \n

    \u817e\u8baf\u6df7\u5143 \/ Z.AI \/ \u667a\u8c31 AI \u7b49\u81ea\u5bb6\u5e73\u53f0<\/span><\/p>\n<\/td>\n

    		\n

    \u5404\u5bb6\u90fd\u6709\u81ea\u5df1\u7684\u5f00\u653e\u5e73\u53f0\u548c\u5f00\u6e90\u6a21\u578b（\u5982 GLM4.x\u3001GLMImage\u3001Hunyuan Image \u7b49），\u90e8\u5206\u4e5f\u540c\u6b65\u5230 HF \u548c ModelScope\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u60f3\u6df1\u5ea6\u4f7f\u7528\u67d0\u5bb6\u5b98\u65b9 API + \u5176\u5bf9\u5e94\u7684\u5f00\u6e90\u6743\u91cd（\u672c\u5730\u5316\u3001\u4e8c\u6b21\u8bad\u7ec3）\u65f6，\u53ef\u4ee5\u76f4\u63a5\u4ece\u8be5\u5bb6\u5b98\u7f51\u548c HF\/ModelScope \u53cc\u7ebf\u83b7\u53d6\u3002<\/span><\/p>\n<\/td>\n

    		\n

    \u5404\u5bb6\u5b98\u7f51 \/ HF \u4e3b\u9875\u53ef\u641c\u5230<\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    3. \u6392\u884c\u699c & \u6bd4\u8f83\u578b\u7f51\u7ad9<\/span><\/h4>\n\n\n\n\n\n\n
    \n

    \u7f51\u7ad9<\/span><\/p>\n<\/td>\n

    		\n

    \u7528\u9014<\/span><\/p>\n<\/td>\n

    		\n

    \u8bf4\u660e<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    Hugging Face Open LLM Leaderboard<\/span><\/p>\n<\/td>\n

    		\n

    \u5f00\u6e90\u5927\u6a21\u578b\u7efc\u5408\u6392\u884c\u699c<\/span><\/p>\n<\/td>\n

    		\n

    \u4ee5\u524d\u662f\u6700\u6743\u5a01\u7684\u5f00\u6e90 LLM \u699c\u4e4b\u4e00，\u6839\u636e MMLUPro\u3001BBH\u3001IFEval \u7b49\u591a\u57fa\u51c6\u7ed9\u51fa\u7efc\u5408\u8bc4\u5206，\u867d\u7136\u73b0\u5df2\u5b58\u6863，\u4f46\u4ecd\u53ef\u4f5c\u4e3a\u5386\u53f2\u53c2\u8003\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    ArtificialAnalysis \/ LLMStats \/ Lambda \/ Vellum Leaderboard \u7b49<\/span><\/p>\n<\/td>\n

    		\n

    LLM \u7efc\u5408\u5bf9\u6bd4<\/span><\/p>\n<\/td>\n

    		\n

    \u63d0\u4f9b MMLU\u3001GPQA\u3001HumanEval\u3001SWEBench \u7b49\u591a\u57fa\u51c6\u6570\u636e，\u4ee5\u53ca\u4ef7\u683c（$\/M tokens）\u3001\u901f\u5ea6\u3001\u4e0a\u4e0b\u6587\u957f\u5ea6\u7b49\u4fe1\u606f，\u53ef\u76f4\u63a5\u8f93\u5165\u6a21\u578b\u540d\u5bf9\u6bd4\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n

    \n

    \u5404\u7c7b\u4e13\u9898\u699c\u5355<\/span><\/p>\n<\/td>\n

    		\n

    OCR\u3001\u4ee3\u7801\u3001\u56fe\u50cf\u751f\u6210\u7b49\u5782\u76f4\u9886\u57df\u699c\u5355<\/span><\/p>\n<\/td>\n

    		\n

    \u5982 GLMImage \u5728\u591a\u4e2a\u6587\u672c\u6e32\u67d3\u57fa\u51c6\u4e0a\u6392\u540d\u524d\u5217\u3001Fish Speech \u5728 TTS Arena \u5f97\u5206\u9760\u524d\u7b49，\u90fd\u53ef\u4ee5\u5728\u76f8\u5e94\u699c\u5355\u7f51\u7ad9\u67e5\u5230\u7ec6\u8282\u3002<\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    \n

    \n

    \u4e94\u3001\u7ed3\u5408\u4f60\u7684\u9700\u6c42\u600e\u4e48\u843d\u5730\u9009\u62e9？<\/span><\/h3>\n

    \u4f60\u539f\u95ee\u9898\u6bd4\u8f83\u7b80\u77ed：\u201c\u5e38\u7528\u5f00\u6e90\u6a21\u578b\u5bf9\u6bd4 + \u5e38\u7528\u5f00\u6e90\u6a21\u578b\u7f51\u7ad9\u201d\u3002\u5982\u679c\u628a\u5b83\u843d\u5230<\/span>\u80fd\u76f4\u63a5\u64cd\u4f5c<\/span>\u7684\u5c42\u9762，\u53ef\u4ee5\u6309\u4e0b\u9762\u51e0\u6b65\u8d70：<\/span><\/p>\n

    \u5148\u5b9a\u65b9\u5411（\u6587\u5b57 \/ \u56fe\u50cf \/ \u8bed\u97f3 \/ \u4ee3\u7801）\u200b<\/span><\/p>\n

      \n
    • \n
        \n
      • \u804a\u5929\u3001\u95ee\u7b54\u3001\u5199\u4f5c：\u770b\u4e0a\u9762 LLM \u8868\u683c\u3002<\/span><\/li>\n
      • \u751f\u56fe：Stable Diffusion 3.5 \/ FLUX \/ GLMImage \/ Hunyuan Image 3.0\u3002<\/span><\/li>\n
      • \u8bed\u97f3：Fish Speech \/ OpenAudioS1 \/ Whisper\u3002<\/span><\/li>\n
      • \u4ee3\u7801\u52a9\u7406：DeepSeekV3.2 \/ Qwen3Coder \/ Llama 4 + \u4e13\u7528 code \u6a21\u578b\u3002<\/span><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n

        \u518d\u9009\u751f\u6001\u5e73\u53f0<\/span><\/p>\n

          \n
        • \n
            \n
          • \u60f3\u8981<\/span>\u5168\u7403\u751f\u6001 & \u82f1\u6587\u8d44\u6599\u591a<\/span>：\u4ece Hugging Face \u641c\u8d77\u3002<\/span><\/li>\n
          • \u4e3b\u8981\u662f<\/span>\u4e2d\u6587\u5f00\u53d1\u8005 \/ \u5728\u56fd\u5185\u670d\u52a1\u5668\u90e8\u7f72<\/span>：\u4f18\u5148 ModelScope，\u518d\u5bf9\u7167 Hugging Face\u3002<\/span><\/li>\n
          • \u4e0d\u60f3\u81ea\u5df1\u8fd0\u7ef4 GPU，\u53ea\u60f3\u7528 API：OpenRouter \/ Together \/ \u963f\u91cc\u4e91 \/ \u767e\u5ea6 \/ \u817e\u8baf\u7b49\u5b98\u65b9\u5e73\u53f0\u3002<\/span><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n

            \u7528\u6392\u884c\u699c\u505a sanity check<\/span><\/p>\n

              \n
            • \n
                \n
              • \u5728 Hugging Face \/ LLMStats \/ ArtificialAnalysis \u770b\u4e00\u4e0b： <\/span>\n
                  \n
                • MMLU \/ MMLUPro：\u901a\u7528\u77e5\u8bc6\u548c\u63a8\u7406<\/span><\/li>\n
                • HumanEval \/ LiveCodeBench：\u4ee3\u7801<\/span><\/li>\n
                • GPQA：\u9ad8\u96be\u5ea6\u95ee\u7b54<\/span><\/li>\n<\/ul>\n<\/li>\n
                • \u540c\u4e00\u68af\u961f\u7684\u6a21\u578b，\u8c01\u66f4\u4fbf\u5b9c\u3001\u8c01\u751f\u6001\u66f4\u6d3b\u8dc3，\u5c31\u4f18\u5148\u8c01\u3002<\/span><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n

                  \u771f\u6b63\u843d\u5730\u65f6\u7684\u4e00\u4e2a\u7b80\u5355\u7ec4\u5408\u793a\u4f8b<\/span><\/p>\n

                    \n
                  • \n
                      \n
                    • \u4e2d\u6587\u4e1a\u52a1 + \u79c1\u6709\u5316\u90e8\u7f72<\/span>： <\/span>\n
                        \n
                      • \u5e73\u53f0：ModelScope + \u672c\u5730 GPU \/ \u79c1\u6709\u4e91<\/span><\/li>\n
                      • LLM：Qwen332B \/ 72B \u6216 DeepSeekV3.2<\/span><\/li>\n
                      • \u751f\u56fe：GLMImage + Stable Diffusion 3.5<\/span><\/li>\n
                      • \u8bed\u97f3：Fish Speech V1.5<\/span><\/li>\n<\/ul>\n<\/li>\n
                      • \u8de8\u56fd\u4e1a\u52a1 \/ \u591a\u8bed\u79cd\u4ea7\u54c1<\/span>： <\/span>\n
                          \n
                        • \u5e73\u53f0：Hugging Face + OpenRouter \/ Together<\/span><\/li>\n
                        • LLM：Llama 4 Maverick + Qwen3 + DeepSeekV3.2（\u6309\u573a\u666f\u5207\u6362）<\/span><\/li>\n
                        • \u751f\u56fe：FLUX.2 Pro \/ Stable Diffusion 3.5<\/span><\/li>\n
                        • \u8bed\u97f3：Fish Speech + Whisper<\/span><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"

                          \u4e00\u3001\u5e38\u7528\u5f00\u6e90\u5927\u8bed\u8a00\u6a21\u578b\u5bf9\u6bd4（LLM）
                          \n1. \u4e3b\u6d41\u901a\u7528 LLM：\u600e\u4e48\u9009？
                          \n\u573a\u666f：\u5bf9\u8bdd\u3001\u5199\u4f5c\u3001\u95ee\u7b54\u3001\u5206\u6790\u3001RAG\u3001\u5927\u90e8\u5206\u901a\u7528\u5e94\u7528\u3002 \u6a21\u578b\u5bb6\u65cf \u4ee3\u8868\u578b\u53f7（\u5f00\u6e90\u6743\u91cd） \u7efc\u5408\u5b9e\u529b & \u7279\u70b9 \u9002\u5408\u4f60\u5728\u4ec0\u4e48\u573a\u666f\u7528 Qwen（\u901a\u4e49\u5343\u95ee，\u963f\u91cc）\u200b Qwen2 \/ Qwen2.5 \/ Qwen3-235B-A22B\u00a0\u7b49 \u5728\u65af\u5766\u798f HELM \/<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[7127,50,132,427],"topic":[],"class_list":["post-67169","post","type-post","status-publish","format-standard","hentry","category-server","tag-7127","tag-50","tag-132","tag-427"],"yoast_head":"\n\u5f00\u6e90AI\u6a21\u578b\u7ec8\u6781\u5bfc\u822a\uff1a\u4eceLLM\u5230\u591a\u6a21\u6001\uff0c\u4e00\u6587\u6559\u4f60\u5982\u4f55\u9009\u578b\u4e0e\u843d\u5730 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wsisp.com\/helps\/67169.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u5f00\u6e90AI\u6a21\u578b\u7ec8\u6781\u5bfc\u822a\uff1a\u4eceLLM\u5230\u591a\u6a21\u6001\uff0c\u4e00\u6587\u6559\u4f60\u5982\u4f55\u9009\u578b\u4e0e\u843d\u5730 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\u4e00\u3001\u5e38\u7528\u5f00\u6e90\u5927\u8bed\u8a00\u6a21\u578b\u5bf9\u6bd4（LLM） 1. \u4e3b\u6d41\u901a\u7528 LLM：\u600e\u4e48\u9009？ \u573a\u666f：\u5bf9\u8bdd\u3001\u5199\u4f5c\u3001\u95ee\u7b54\u3001\u5206\u6790\u3001RAG\u3001\u5927\u90e8\u5206\u901a\u7528\u5e94\u7528\u3002 \u6a21\u578b\u5bb6\u65cf \u4ee3\u8868\u578b\u53f7（\u5f00\u6e90\u6743\u91cd） \u7efc\u5408\u5b9e\u529b & \u7279\u70b9 \u9002\u5408\u4f60\u5728\u4ec0\u4e48\u573a\u666f\u7528 Qwen（\u901a\u4e49\u5343\u95ee，\u963f\u91cc）\u200b Qwen2 \/ Qwen2.5 \/ Qwen3-235B-A22B\u00a0\u7b49 \u5728\u65af\u5766\u798f HELM \/\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/67169.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-01-28T05:06:51+00:00\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/67169.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/67169.html\",\"name\":\"\u5f00\u6e90AI\u6a21\u578b\u7ec8\u6781\u5bfc\u822a\uff1a\u4eceLLM\u5230\u591a\u6a21\u6001\uff0c\u4e00\u6587\u6559\u4f60\u5982\u4f55\u9009\u578b\u4e0e\u843d\u5730 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-01-28T05:06:51+00:00\",\"dateModified\":\"2026-01-28T05:06:51+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/67169.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/67169.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/67169.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u5f00\u6e90AI\u6a21\u578b\u7ec8\u6781\u5bfc\u822a\uff1a\u4eceLLM\u5230\u591a\u6a21\u6001\uff0c\u4e00\u6587\u6559\u4f60\u5982\u4f55\u9009\u578b\u4e0e\u843d\u5730\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u5f00\u6e90AI\u6a21\u578b\u7ec8\u6781\u5bfc\u822a\uff1a\u4eceLLM\u5230\u591a\u6a21\u6001\uff0c\u4e00\u6587\u6559\u4f60\u5982\u4f55\u9009\u578b\u4e0e\u843d\u5730 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wsisp.com\/helps\/67169.html","og_locale":"zh_CN","og_type":"article","og_title":"\u5f00\u6e90AI\u6a21\u578b\u7ec8\u6781\u5bfc\u822a\uff1a\u4eceLLM\u5230\u591a\u6a21\u6001\uff0c\u4e00\u6587\u6559\u4f60\u5982\u4f55\u9009\u578b\u4e0e\u843d\u5730 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\u4e00\u3001\u5e38\u7528\u5f00\u6e90\u5927\u8bed\u8a00\u6a21\u578b\u5bf9\u6bd4（LLM） 1. \u4e3b\u6d41\u901a\u7528 LLM：\u600e\u4e48\u9009？ \u573a\u666f：\u5bf9\u8bdd\u3001\u5199\u4f5c\u3001\u95ee\u7b54\u3001\u5206\u6790\u3001RAG\u3001\u5927\u90e8\u5206\u901a\u7528\u5e94\u7528\u3002 \u6a21\u578b\u5bb6\u65cf \u4ee3\u8868\u578b\u53f7（\u5f00\u6e90\u6743\u91cd） \u7efc\u5408\u5b9e\u529b & \u7279\u70b9 \u9002\u5408\u4f60\u5728\u4ec0\u4e48\u573a\u666f\u7528 Qwen（\u901a\u4e49\u5343\u95ee，\u963f\u91cc）\u200b Qwen2 \/ Qwen2.5 \/ Qwen3-235B-A22B\u00a0\u7b49 \u5728\u65af\u5766\u798f HELM \/","og_url":"https:\/\/www.wsisp.com\/helps\/67169.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-01-28T05:06:51+00:00","author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"4 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/67169.html","url":"https:\/\/www.wsisp.com\/helps\/67169.html","name":"\u5f00\u6e90AI\u6a21\u578b\u7ec8\u6781\u5bfc\u822a\uff1a\u4eceLLM\u5230\u591a\u6a21\u6001\uff0c\u4e00\u6587\u6559\u4f60\u5982\u4f55\u9009\u578b\u4e0e\u843d\u5730 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-01-28T05:06:51+00:00","dateModified":"2026-01-28T05:06:51+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/67169.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/67169.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/67169.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"\u5f00\u6e90AI\u6a21\u578b\u7ec8\u6781\u5bfc\u822a\uff1a\u4eceLLM\u5230\u591a\u6a21\u6001\uff0c\u4e00\u6587\u6559\u4f60\u5982\u4f55\u9009\u578b\u4e0e\u843d\u5730"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/67169","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=67169"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/67169\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=67169"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=67169"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=67169"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=67169"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}