How to switch model and permissions inline with /model and /permissions in Codex CLI

Starting with Codex CLI 0.164.0, /model and /permissions accept arguments. You can change the model, reasoning effort, speed or permission profile in one line without opening a picker. Bare /model still opens the picker.

1. Change model, reasoning effort and speed together. Leave out the model to change settings for the current one.

/model gpt-6.1-sol --reasoning high
/model --reasoning high --speed standard
/model --speed=fast

Both --reasoning high and --reasoning=high forms work. Speed is standard or fast. Bare values after the model, like high fast, are rejected, so always use the flags. Models your account can't use, or reasoning levels and speeds the model doesn't support, are refused. These changes apply to the current session only and leave your saved defaults untouched.

1. Pick a permission profile by name.

/permissions :read-only
/permissions :ask-for-approval

The aliases :ask-for-approval and :approve-for-me work, and custom profiles can be selected by name too. :approve-for-me only works where that feature is enabled. Switching to full access still asks for confirmation, and queued messages wait until it's confirmed. An unknown name tells you to run /permissions to see available profiles.

Put lines like these in /presets slots to switch modes with one key. This isn't in the docs yet; it comes from the changelog and the repository's pull requests.