opencode with LM Studio

Updated 27 May 2026

Run opencode using local models served by LM Studio.

Setup

1. Configure LM Studio

• Download and install LM Studio
• Load a model (e.g. Gemma 3n-e4b, Qwen 2.5 Coder 14b)
• Start the local server: Developer tab → Start Server (default: http://127.0.0.1:1234)
• Set context length: Settings → Model Defaults → Default Context Length → Model Maximum

2. Copy the config

cp opencode.json ~/.config/opencode/opencode.json

3. Run opencode

opencode

Once opencode is open, type /model to switch to a model prefixed with LM Studio (local).

Note: Running large models locally is often too slow for practical coding use. For a cheaper alternative to Claude/GPT, the DeepSeek API offers competitive performance at a fraction of the cost.

Config

opencode.json points opencode at the LM Studio server and registers local models. Edit the models block to add or remove models — the key must match the model ID shown in LM Studio.

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "lmstudio": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "LM Studio (local)",
      "options": {
        "baseURL": "http://127.0.0.1:1234/v1"
      },
      "models": {
        "google/gemma-3n-e4b": {
          "name": "Gemma 3n-e4b (local)"
        },
        "qwen/qwen2.5-coder-14b" : {
          "name": "Qwen 2.5 Coder 14b (local)"
        }
      }
    }
  }
}