Codex is dumber and also consume a lot of tokens
Open 💬 3 comments Opened Jun 9, 2026 by nutmiguelelias-blip
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of the Codex App are you using (From “About Codex” dialog)?
26.601.10930
What subscription do you have?
Pro x5
What platform is your computer?
_No response_
What issue are you seeing?
Lately, it has been giving very limited responses, and it also seems to be struggling more with the same tasks. Even with Skills and MD Agents, it often fails to complete them, and I have to resend the prompt. It can't activate extensions, it can't use my computer properly, and although I can sometimes force it to complete the task, it still runs into the same type of error. It feels a bit less capable overall, and on top of that, it consumes even more tokens.
What steps can reproduce the bug?
Feedback ID: 019ea6e5-9bac-7230-8b41-f3cfde73eff6
What is the expected behavior?
_No response_
Additional information
_No response_
3 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
I have noticed that for the past two weeks (since this issue was filed) codex has been a lot less capable. Seems to happen when openai (or anyone else, noticed it with claude when I used it) is preparing a new model to launch. My guess is they need extra compute so they throttle current models a little bit, some users notice the degradation but can't prove anything so why wouldn't they?
same