Select model for plan implementation
Resolved 💬 5 comments Opened Mar 17, 2026 by curious1n Closed Aug 6, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What variant of Codex are you using?
App
What feature would you like to see?
In Plan mode in Codex App when I click on "Implement this" - I'd like to change the model (eg. Plan in xhigh and switch to high for implementation) - Not sure if Codex CLI / config.toml has that support, btw.
Additional information
_No response_
5 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
For now I handle this by just declining the "implement plan" prompt and then setting the effort level and sending a message "Implement plan.". It has the same outcome as what you want, but it is chunky and you have to remember that you want to change the reasoning effort.
On my best days I handle this the way day @GregPeden does. But more often I forget to do that, so lots of extra work is generated because once the implementation starts on extra high I'm hesitant to change it. I imagine this is good neither for OpenAI's servers nor my credits, so it would be nice if there was a way to configure differing defaults for plan and implement to begin with; and an input to change reasoning level and model at time of switching from plan to implementation; and for good measure a more clear way to course correct if implementation started on the wrong level (the UI control allows changes at any time, it's just not clear to me as a user if they are doing anything or what the risks may be of doing so.)
Strong support. The key missing piece is the handoff moment from Plan -> Implement.
When I click “Implement this”, I should be able to confirm:
Right now the workaround is manual and easy to miss, which leads to accidental high-cost runs. A small handoff selector would fix most of the UX pain without adding much complexity.
I have seen it implemented in both Windows and macOS recently. Three is a dropdown button on the left side under the question prompt allowing the user to choose model and effort.