Massive CPU Usage in macOS app
What version of Codex is running?
Version 260202.0859 (517)
What subscription do you have?
Plus
Which model were you using?
Codex 5.2 xhigh
What platform is your computer?
Darwin 25.2.0 arm64 arm
What terminal emulator and version are you using (if applicable)?
iTerm2
What issue are you seeing?
CPU goes wild doing pretty much any task, but especially plan tasks with many files and searches.
<img width="1926" height="204" alt="Image" src="https://github.com/user-attachments/assets/01c35ae0-a123-40a4-a154-2aa3d5cb0a3b" />
Quite impressive, Codex has made my MacBook get hot for the first time since I got it 2 years ago. Never heard the fans until today. Not Docker, PhpStorm, Chrome, .... but Codex managed to burn my lap. Impressive. I apologize for the snark but I am genuinely curious how ONE Codex session can cause this while +5 concurrent Claude sessions (in their crappy React TUI) doesn't.
What steps can reproduce the bug?
Start any session, eg plan or a larger feature request. Check CPU usage.
What is the expected behavior?
Lap not burnt.
Additional information
_No response_
12 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Duplicate of #10432
@etraut-openai I'd suggest reopening this issue as this seems to be connected to a different problem compared to #10432.
In my case, I'm not seeing any spikes in GPU, but CPU consistently hitting >100% when the chat is streaming _and_ the diff view is open, especially as diffs grow larger.
I'm assuming that the diff view is being rerendered when the chat updates, likely something that could be fixed by either memoization or making sure that the shared parent doesn't rerender when the chat itself updates.
Yes, I agree this isn't the same as the GPU spiking issue. Reopening.
We think we understand most of the high-CPU-utilization problem, although there are multiple contributors. We are testing fixes for the contributors that we understand and should have a fix out shortly.
@etraut-openai I think this has been fixed in Version 260210.1703 (602)?
Speaking tentatively based on the last hour of usage.
@Suhaibinator, thanks for the tentative confirmation. The team fixed a number of performance-related issues over the past couple of days. It's possible there are still some lurking, but we're hopeful that most of them are now addressed.
Please keep the feedback and reports coming!
Thank you guys for the awesome work you put into this product and especially for addressing the performance issues <3
Battery performance has NOTICEABLY improved. I was burning through about 2% of battery per minute, and now it's at around 1% every ~3-4 minutes (probably a 5-10x improvement)
Great product. Really looking forward to using this on more platforms if/when you choose to create codex for linux. Really loving codex.
I did not see spike in CPU usage but had a very terrible performance issue.
The app has been really slow for the past few days. It's noticeably sluggish when switching sessions, typing, and displaying messages
I am on pre-release (ie no fix up to the time of this comment has changed this) and it's effectively unusable because of the CPU usage. Is there a known good version I can roll back to?
I already submitted feedback. I've seen tickets with this affecting Windows, MacOS, and Linux at this point.
@ppoage there is no codex app on linux...
Still this issue, it heats up even when casual chatting
atm need to disable the extension’s local-environment and app-target integration paths
https://github.com/openai/codex/issues/16231#issuecomment-4205506808