Broken Context Count
Resolved 💬 3 comments Opened Aug 11, 2025 by mcqx4 Closed Oct 18, 2025
💡 Likely answer: A maintainer (tibo-openai, collaborator)
responded on this thread — see the highlighted reply below.
What version of Codex is running?
Last version 11.08.2025
Which model were you using?
gpt-5
What platform is your computer?
mac os m1 pro (2021)
What steps can reproduce the bug?
Critical bug: When the system reaches usage limits, it becomes impossible to continue from where you left off in your coding session. Even after waiting for the limit to reset, entering any prompt results in no request being sent to the AI. This forces users to either start from scratch or restart their terminal.
This issue is particularly severe for API KEY users, as Tier 1 TPM (tokens per minute) limits for GPT-5 are hit frequently during normal usage. Each time the limit is reached, the system completely freezes with no way to resume - it simply stops sending requests to the AI even after the usage limit period has passed.
What is the expected behavior?
_No response_
What do you see instead?
_No response_
Additional information
_No response_
3 Comments
Yes, nasty and annoying bug.
And a big waste of tokens since work has to be repeated.
Codex CLI is virtually unusable for me because of this exact issue.
We now have
codex resume, closing.