Allow hiding / disabling unused models in Codex app and Codex CLI

Resolved 💬 2 comments Opened Mar 8, 2026 by hlemonnier Closed Mar 8, 2026

What variant of Codex are you using?

App (Macos), IDE

What feature would you like to see?

Summary

Please add a way to choose which models are shown in the model selector, both in the Codex app and in Codex CLI.

Right now, several models are always displayed. In practice, I only use a subset of them, so having all models visible adds unnecessary clutter and makes the selector less convenient to use.

Problem

Currently, the model list appears to be fixed. Even when some models are never used, they still appear in the UI / selector.

Proposed solution

Add a way for users to enable or disable models from the list of available models, so that disabled models do not appear in the model selector.

Two possible approaches would work well:

Option 1: settings in the Codex app

Add a settings page or preferences section where users can toggle models on/off.

For example:

  • gpt-x → enabled
  • gpt-y → disabled
  • gpt-z → enabled

Only enabled models would appear in the model selector.

Option 2: configuration via config.toml

Allow the same behavior from configuration, especially for Codex CLI and for users who prefer a config-first workflow.

For example, something like:

[models]
disabled = ["model-d", "model-e", "model-f"]

Then only the active models would be shown in the selector.

Scope

This would ideally apply to:

  • Codex app
  • Codex CLI

And the behavior should be consistent across both.

Expected behavior

If a model is disabled by the user:

  • it should not appear in the model selector
  • it should remain available internally if needed, but hidden from normal selection UI
  • the configuration should persist across sessions

Additional information

_No response_

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗