Choosing a Model

Choosing a Model

Codex Assist sends requests to the Codex backend using a model slug. The goal is to make model selection easy without trapping advanced users when OpenAI changes available models.

Model selector behavior

In versions with the model selector:

• During first setup, Codex Assist shows a curated fallback list because the account is not authenticated yet.
• In the options flow, Codex Assist tries to load available models from the Codex backend for the signed-in account.
• Hidden backend models are filtered out.
• Backend priority is used when available.
• Known Codex-compatible fallback slugs remain available if discovery fails.
• Custom values are still supported for advanced users.

Why keep custom values?

Codex model availability can change upstream. If a new model slug works before the discovery list catches up, custom values let you test it without waiting for a new release.

Recommended approach

Most users should pick from the dropdown and leave the prompt alone at first.

Advanced users can change the model when they want to compare speed, cost/usage behavior, or response quality across available Codex models.

If a model stops working

Try this order:

1. Open Codex Assist options and select another listed model.
2. Save options and test a simple Assist prompt.
3. If all models fail with auth errors, complete reauth from Home Assistant Repairs or the integration page.
4. If the backend appears changed, check the project issues and latest release notes.