gpt 5.1-codex-max xhigh refuses task because it would take too long (lol)
What version of Codex is running?
0.63
What subscription do you have?
PRO
Which model were you using?
gpt-5.1-codex-max (reasoning xhigh, summaries auto)
What platform is your computer?
WSL ubuntu
What issue are you seeing?
codex-5.1-max xhigh basically refuses because the task would be too long and would not fit in a single iteration the screenshot will be in portuguese but i asked it to take a huge test suite and start writing the code to pass that suite it refused and even says there is no possibility of compressing itself along the way
What steps can reproduce the bug?
it is not possible to reproduce the exact scenario without being in my repo but basically i took a huge test suite in node and told it to start implementing the code that would pass the suite
What is the expected behavior?
well the codex max system card says it can work for hours because of the optimization they made in its ability to self compress after it refused because it explicitly said the request would not fit in a single iteration i told it you can self compress along the way and it replied saying there is no possibility of self compressing
<img width="1889" height="379" alt="Image" src="https://github.com/user-attachments/assets/6c818c8b-57c6-4a44-9a25-e3266d246c25" />
<img width="1908" height="382" alt="Image" src="https://github.com/user-attachments/assets/a76fa944-16d7-4c73-9c8d-5a317ed4d109" />
it explicitly says this would take hours or days i cannot do it
Additional information
_No response_
4 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
it is not a duplicate in the other reports codex at least starts the implementation in mine it totally refuses because it thinks it would take too long
This problem existed long before gpt-5.1-codex-max xhigh because when it gets too complicated, it stops the task because it could get it completely wrong. I think a sort of forced ‘warning’ barrier has been put in place because if you continued, it could cause significant damage to your repository.
Thanks for the bug report. We'll share the feedback with the team responsible for training our codex models. If you see other model behavior that you'd like to report, please use the
/feedbackcommand. That will provide us with additional details including logs and conversation text.