codex - 💡(How to fix) Fix Codex Pro GPT-5.4 long context regressed: ~672k worked Apr 22, ~290k fails Apr 23 [8 comments, 6 participants]

TL;DR

The issue is likely caused by a mismatch between the client's auto-compact limit and the backend's context window limit, resulting in the Codex app rejecting requests around the 272k token boundary.

Guidance

• Verify that the model_auto_compact_token_limit is being correctly sent to the backend and that it is being respected.
• Check the backend logs to see if there are any errors or warnings related to the context window limit or auto-compact limit.
• Consider reducing the model_auto_compact_token_limit to a value closer to the 272k boundary to see if it resolves the issue.
• Investigate if the upgrade to codex-cli 0.123.0 introduced any changes to the context window limit or auto-compact behavior.

Example

No code snippet is provided as the issue seems to be related to configuration and backend behavior.

Notes

The issue may be specific to the GPT-5.4 model or the pro plan, and further investigation is needed to determine the root cause.

Recommendation

Apply workaround: Reduce the model_auto_compact_token_limit to a value closer to the 272k boundary to see if it resolves the issue, as the current limit seems to be mismatched with the backend's context window limit.