环境Win11opencode有免费模型DeepSeek V4 Flash Free (New)可以用这个应该挺好用的。一、相关链接Open Code 官网OpenCode | 开源 AI 编程代理Open Code 文档下载 OpenCode[1] OpenCode 终端[2] OpenCode 桌面版[3] OpenCode 扩展[4] OpenCode 集成githubopencode/README.zh.md at dev · anomalyco/opencode · GitHub二、前置条件curl http://localhost:11434/v1/models # 能返回模型列表说明服务正常Windows 桌面版 Ollama 启动后会常驻后台并自动监听 11434 端口ollama serve 启动的方便看日志三、连接本地模型Ollama3.1 Ollama一键启动指定模型ollama launch opencode --model qwen3.6:35b会自动修改配置并启动 OpenCode适合快速试用。不会覆盖已有的阿里云百炼配置。C:\Users\adminollama launch opencode --model qwen3.6:35b This will modify your OpenCode configuration: C:\Users\admin\.local\state\opencode\model.json Backups will be saved to C:\Users\admin\AppData\Local\Temp\ollama-backups/ Proceed? Yes No ←/→ navigate • enter confirm • esc cancel 选Yes如果上次用的是这个模型就不用选了如果Windows下没有运行以ollama serve启动的服务他会把Ollama应用拉起来作为后台服务。首页​聊天​退出/exit​交互式选择模型ollama launch opencode​3.2 添加自定义提供商opencode.jsonc修改配置文件opencode.jsoncC:\Users\admin\.config\opencode\opencode.jsoncmodels 里的 key 必须和 ollama list 显示的模型名完全一致CLI 和桌面版读的是同一份配置配好一处两端通用。{ $schema: https://opencode.ai/config.json, model: opencode/deepseek-v4-flash-free, provider: { ollama: { npm: ai-sdk/openai-compatible, name: Ollama (local), options: { baseURL: http://localhost:11434/v1 }, models: { qwen3.6:27b: { name: Qwen3.6 27B }, qwen3.6:35b: { name: Qwen3.6 35B } } } } }带limit的{ $schema: https://opencode.ai/config.json, model: opencode/deepseek-v4-flash-free, provider: { ollama: { npm: ai-sdk/openai-compatible, name: Ollama (local), options: { baseURL: http://localhost:11434/v1 }, models: { qwen3.6:27b: { name: Qwen3.6 27B, limit: { context: 65536, output: 8192 } }, qwen3.6:35b: { name: Qwen3.6 35B, limit: { context: 65536, output: 8192 } } } } } }查看上下文长度可以查看C:\Users\adminollama psNAME ID SIZE PROCESSORCONTEXTUNTILqwen3.6:35b 07d35212591f 36 GB 100% GPU 262144 4 minutes from now可以查看ollama启动日志default_num_ctxOLLAMA_CONTEXT_LENGTH:00 表示让 Ollama 自动算default_num_ctx262144 Ollama 根据你的总 VRAM 自动选的默认值OLLAMA_CONTEXT_LENGTH 环境变量名控制全局默认上下文长度。可选配置C:\Users\adminollama show qwen3.6:35bC:\Users\adminollama show qwen3.6:35b Model architecture qwen35moe parameters 36.0B context length 262144 embedding length 2048 quantization Q4_K_M Capabilities completion vision tools thinking Parameters top_k 20 top_p 0.95 min_p 0 presence_penalty 1.5 repeat_penalty 1 temperature 1 License Apache License Version 2.0, January 2004 ...切换模型/models通过opencode启动的切换模型时提供商显示的是Ollama (local)在opencode.jsonc配了自定义提供商切换模型时才显示Ollama的切换模型/modelsOllama启动通过Ollama启动的opencode切换模型时提供商显示的是Ollama3.3 桌面版添加自定义提供商桌面版操作填的内容和之前基本一样opencode.jsoncC:\Users\admin\.config\opencode\opencode.jsonc桌面版的操作会修改配置文件和手动配置基本差不多提供商的name和npm顺序换了一下多了个节点disabled_providers: []{ $schema: https://opencode.ai/config.json, model: opencode/deepseek-v4-flash-free, provider: { ollama: { name: Ollama (local), npm: ai-sdk/openai-compatible, options: { baseURL: http://localhost:11434/v1 }, models: { qwen3.6:27b: { name: Qwen3.6 27B }, qwen3.6:35b: { name: Qwen3.6 35B } } } }, disabled_providers: [] }