https://turingparty-ai.pages.dev/tags/%E5%BC%80%E6%BA%90%E6%A8%A1%E5%9E%8B/Recent content in 开源模型 on 图灵派对Hugozh-cnWed, 25 Mar 2026 00:00:00 +0000https://turingparty-ai.pages.dev/posts/20260325/Wed, 25 Mar 2026 00:00:00 +0000https://turingparty-ai.pages.dev/posts/20260325/<p>不想把数据发送到第三方 API？想在自己的机器上跑大模型？Ollama + Open WebUI 是目前最简单的本地部署方案。</p> <h2 id="为什么要本地部署">为什么要本地部署？</h2> <ul> <li><strong>隐私保护</strong>：数据不出本地</li> <li><strong>零成本</strong>：无需 API 费用</li> <li><strong>低延迟</strong>：无网络往返</li> <li><strong>可定制</strong>：自由选择和切换模型</li> </ul> <h2 id="ollama-安装">Ollama 安装</h2> <div class="highlight"><pre tabindex="0" style="color:#cdd6f4;background-color:#1e1e2e;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"><code class="language-bash" data-lang="bash"><span style="display:flex;"><span><span style="color:#6c7086;font-style:italic"># macOS / Linux</span> </span></span><span style="display:flex;"><span>curl -fsSL https://ollama.com/install.sh | sh </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span><span style="color:#6c7086;font-style:italic"># Windows</span> </span></span><span style="display:flex;"><span><span style="color:#6c7086;font-style:italic"># 从 https://ollama.com/download 下载安装包</span> </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span><span style="color:#6c7086;font-style:italic"># 拉取模型</span> </span></span><span style="display:flex;"><span>ollama pull llama3.1:8b </span></span><span style="display:flex;"><span>ollama pull qwen2.5:7b </span></span></code></pre></div><h2 id="open-webui-部署">Open WebUI 部署</h2> <div class="highlight"><pre tabindex="0" style="color:#cdd6f4;background-color:#1e1e2e;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"><code class="language-bash" data-lang="bash"><span style="display:flex;"><span><span style="color:#6c7086;font-style:italic"># Docker 一键部署</span> </span></span><span style="display:flex;"><span>docker run -d -p 3000:8080 <span style="color:#89b4fa">\ </span></span></span><span style="display:flex;"><span> --add-host<span style="color:#89dceb;font-weight:bold">=</span>host.docker.internal:host-gateway <span style="color:#89b4fa">\ </span></span></span><span style="display:flex;"><span> -v open-webui:/app/backend/data <span style="color:#89b4fa">\ </span></span></span><span style="display:flex;"><span> --name open-webui <span style="color:#89b4fa">\ </span></span></span><span style="display:flex;"><span> ghcr.io/open-webui/open-webui:main </span></span></code></pre></div><p>访问 <code>http://localhost:3000</code> 即可使用类 ChatGPT 的界面与本地模型对话。</p> <h2 id="硬件要求">硬件要求</h2> <ul> <li><strong>7B 模型</strong>：8GB 内存 / 6GB 显存</li> <li><strong>13B 模型</strong>：16GB 内存 / 10GB 显存</li> <li><strong>70B 模型</strong>：64GB 内存 / 48GB 显存</li> </ul> <blockquote> <p>没有 GPU 也能跑！Ollama 支持 CPU 推理，7B 模型在 M1 Mac 上体验流畅。</p>