Configure GPT models in Claude Code

Claude Code can call GPT models through the LMU AI API, for example gpt-5.6-sol, gpt-5.6-terra, and gpt-5.5.

Config file location

Claude Code's config file is settings.json.

~/.claude/settings.json

C:\Users\YourUsername\.claude\settings.json

Option 1 — Full config (recommended)

The full config maps Claude Code's internal Opus / Sonnet / Haiku default models to GPT models. Once configured, launching Claude Code uses the corresponding GPT model by default.

Open settings.json in a text editor and add the following (replace the key with your own LMU AI API Key):

{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "sk-your-lmu-ai-api-key",
    "ANTHROPIC_BASE_URL": "https://api.lmuai.ai",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "gpt-5.4",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL_NAME": "gpt-5.4",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "gpt-5.6-sol",
    "ANTHROPIC_DEFAULT_OPUS_MODEL_NAME": "gpt-5.6-sol",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "gpt-5.5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL_NAME": "gpt-5.5"
  },
  "skipDangerousModePermissionPrompt": true,
  "model": "opus"
}

Field reference:

Option 2 — Minimal config

If you just want to configure the key and Base URL first, you can use the minimal config:

{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "sk-your-lmu-ai-api-key",
    "ANTHROPIC_BASE_URL": "https://api.lmuai.ai"
  },
  "skipDangerousModePermissionPrompt": true,
  "model": "opus"
}

The minimal config doesn't map opus to gpt-5.6-sol ahead of time. After launching Claude Code, run this manually in Claude Code:

/model gpt-5.6-sol

The current session then switches to gpt-5.6-sol.

Common issues

Should the Base URL include /v1?

Claude Code connects to the LMU AI API using the Anthropic protocol, so use:

https://api.lmuai.ai

Do not use https://api.lmuai.ai/v1.

Can I just set model to gpt-5.6-sol?

We recommend one of two approaches instead:

• Use the full config, mapping ANTHROPIC_DEFAULT_OPUS_MODEL to gpt-5.6-sol and keeping "model": "opus"
• Use the minimal config and run /model gpt-5.6-sol after launch

Both fit Claude Code's model-tier configuration approach better, and make switching and reverting clearer.

What if it still doesn't work after configuring?

Check the following:

1. Whether the API Key is a key generated in the LMU AI console
2. Whether ANTHROPIC_BASE_URL is https://api.lmuai.ai
3. Whether you've subscribed to a GPT plan, or, on pay-as-you-go, selected the GPT group
4. Whether the model name is spelled correctly, e.g. gpt-5.6-sol

To see the currently supported models, check the Model Gallery.

Last updated: 2026-08-26

Still have questions? Ask in the LMU AI Community (in Chinese) and sign in with your console account.