Using LlamaBarn with OpenCode

[erusev]

01 — Configure LlamaBarn

• Ensure LlamaBarn is running.
• Install the model(s) you want to use with Claude Code.
• Go to Settings and set the Context length to 32k tokens or more.

02 — Configure OpenCode

Create ~/.config/opencode/opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "LlamaBarn": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "http://localhost:2276/v1"
      },
      "models": {
        "gpt-oss-20b": {},
        "qwen3-coder": {}
      }
    }
  }
}

Replace gpt-oss-20b and qwen3-coder with the IDs of the models you want to use (right-click on a model to get its ID).

03 — Run OpenCode and select a LlamaBarn model

Type /models to bring up the model selection menu, then choose a LlamaBarn model.

Recommended models

• gpt-oss-20b — Requires Mac with 24GB+ memory
• gpt-oss-120b — Requires Mac with 96GB+ memory
• qwen3-coder — Requires Mac with 32GB+ memory