Context window at 42% left, complaining "Your input exceeds the context window of this model"

What version of Codex is running?

0.29.0

Which model were you using?

GPT5-high

What platform is your computer?

macOS

What steps can reproduce the bug?

<img width="1636" height="562" alt="Image" src="https://github.com/user-attachments/assets/c7882da7-bda0-4a8a-afab-2fe8d50e977c" />

This happens every time I'm near or around 40% any new input is refused and it complains about context window size. Why is it saying 40% left, is that a bug in how it's counting tokens or is that a bug elsewhere? Many times the input that triggers this is just a continuation and the change requested is minimal, nothing that would consume 40% of remaining tokens.

What is the expected behavior?

40% is a lot of tokens remaining I expect it to continue to work

What do you see instead?

Complains input exceeeds context window size.

Additional information

_No response_