Taking a long time / doesn't answer

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

• #6519
• #6280
• #5885
• #6331
• #6277

Powered by Codex Action

same

Same here with a Pro account. It just sits there doing nothing for almost 14 minutes before beginning work at a seemingly normal pace. It's fundamentally the experience you'd get if someone's preferred method of debugging backend components that can't just have a debugger attached for their entire execution lifetime was to a log a message somewhere and spin in place for long enough to manually attach a debugger, and their preferred duration to wait was 13:37 because they thought it was funny.

I downgraded (npm install -g @openai/codex@0.57.0), it's working now and I still have 5.1 models.

Same here with a pro account.

Same here. It gets stuck after reading a file or two. Long thinking time, no context usage. Tested with an OpenAPI key.
The issue persists in 0.59.0-alpha.5 as well.

Same issue, it's completely unusable at the moment (using codex from its early releases).

Using gpt-5.1-codex high with cli v0.58.0. It's not just an impression it really gets stuck with no replies, and if you escape/reprompt it sometimes completes the steps successfully in under a minute. You're essentially playing Russian roulette and sometimes it works, sometimes it doesn't. It's a frustrating experience.

_edit_: downgrading to v0.57.0 seems to be working so far as suggested in a comment above. So it looks like a regression on v0.58.0 and above.

Same here，stuck at 0.57 for now, I can only use gpt-5.1-codex-max high. I cannot use xhigh until problem was fixed

Issue still persists with ~v0.60.1~ v0.63.0. The first session went well, then on the second session it got stuck 20 minutes on the same step which should have been fast and I had to stop it. On the next try, it completed the same step/task in 55 seconds.

Facing the same issue. Indeed downgrading to 0.57 makes it work.

Same issue, it's completely unusable at the moment (using codex from its early releases). Using gpt-5.1-codex high with cli v0.58.0. It's not just an impression it really gets stuck with no replies, 和 if you escape/reprompt it sometimes completes the steps successfully in under 1 分钟之前. You're essentially playing Russian roulette and sometimes it works, sometimes it doesn't. It's a frustrating experience. _edit_: downgrading to v0.57.0 seems to be working so far as suggested in a comment above. So it looks like a regression on v0.58.0 and above.

Before downgrading to v0.57, all tasks typically took at least half an hour to complete, and in some cases even several hours; after downgrading to version v0.57, under the same model, the task completion time was reduced to approximately 2~10 minutes.

This bug report hasn't received any upvotes or additional posts in three months. Closing.