Extremely low performance on simple tasks
Resolved 💬 8 comments Opened Dec 25, 2025 by Ksizd Closed Jan 6, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of Codex is running?
codex-cli 0.77.0
What subscription do you have?
pro
Which model were you using?
gpt-5.2-codex xhigh
What platform is your computer?
Windows 11 x86-64
What issue are you seeing?
Extremely low performance even in simple tasks! Check git, push git -> thinking for 50 minutes! This is happening to me for 7-10 days already. Unusable
What steps can reproduce the bug?
Run any request and wait for eternity
What is the expected behavior?
Thinking fast for basic tasks (not 1 hour for git --sb)
Additional information
_No response_
8 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Please use the
/feedbackslash command to upload your logs and session details, then post the thread ID here.I tried to use /feedback function, but it stuck after describing bug. Found those .log and .jsonl files manually in file system. I cleared entire log because it was 1 GB already! Made 1 simple request to Codex to show the bug and exported tui log and chat history.
https://limewire.com/d/dfL2A#RdqRaOaSST
Uploaded history
codex-tui.log
@Ksizd, it looks like you're using the
xhighreasoning effort. This will be relatively slow. We recommend "medium" reasoning effort for most coding tasks and "high" for complex tasks like major refactors. For simple tasks, you can typically use "low". The "xhigh" reasoning effort should be used in rare cases for the most complex of tasks.@etraut-openai I face almost the same issues with other reasoning efforts. May the trouble be in internet speed (5 MBit/s)?
A slow internet connection will definitely contribute to longer times.
Here's another thing to try: disable the "undo" feature. Add the following to your
config.tomlfile.This disables an experimental feature that can sometimes lead to major performance degradation. We are planning to remove it for that reason. I see some lines in your logs that lead me to believe that you may be affected by this.
stuck every task
ex:
XXXXXX (1h 45m 28s • esc to interrupt)
We've removed the experimental
/undofeature in the latest version of the CLI, so you no longer need to disable it.