How to use the Pi coding agent with Unquant

Pi supports OpenAI-compatible providers natively. One entry in its models file and Pi runs on Unquant.

Setup guide2 min readUpdated 1 October 2026

1
Set your key
```
export UNQUANT_API_KEY=your-unquant-key
```

Add the provider

Edit ~/.pi/agent/models.json:

{
  "providers": {
    "unquant": {
      "baseUrl": "https://api.unquant.io/v1",
      "api": "openai-completions",
      "apiKey": "$UNQUANT_API_KEY",
      "models": [
        { "id": "glm-5.3-flash", "name": "GLM-5.3-Flash", "contextWindow": 262144 },
        { "id": "qwen-3.8-flash-next", "name": "Qwen 3.8 Flash Next", "contextWindow": 262144 }
      ]
    }
  }
}

3
Run Pi
```
pi --model unquant/glm-5.3-flash
```

Keep reading

How to add Unquant to OpenCodeAdd Unquant to OpenCode as an OpenAI-compatible provider in opencode.json and run the terminal coding agent on GLM-5.3-Flash with 262K context and unlimited tokens.Setup guide · 2 min
How to use Claude Code with an uncensored, unlimited modelRun Claude Code on GLM-5.3-Flash or Qwen 3.8 Flash Next through Unquant's Anthropic-compatible endpoint: two environment variables, no proxy, unlimited tokens and 262K context.Setup guide · 3 min
How to use Reasonix with UnquantRun Reasonix, the DeepSeek terminal coding agent, on Unquant: one provider entry in ~/.reasonix/config.toml and your key in ~/.reasonix/.env.Setup guide · 2 min