codex - 💡(How to fix) Fix GPT-5.5 in Codex ignores configured context window and appears hard-capped around 258K tokens [1 comments, 2 participants]

TL;DR

The effective context limit of around 258,000 tokens for GPT-5.5 in the Codex App may be a hardcoded or technical limitation that prevents the configured 1,000,000-token setting from being fully utilized.

Guidance

• Investigate if there are any technical or performance reasons why the context window is capped at 258,000 tokens, such as memory constraints or processing power limitations.
• Check the Codex App documentation or support resources to see if there are any known limitations or workarounds for this issue.
• Consider testing the same model and settings on a different platform or environment to determine if the issue is specific to the current setup.
• Review the system resources (e.g., RAM, CPU) to ensure they are sufficient to handle the configured context window size.

Notes

The issue may be related to the underlying architecture or technical constraints of the Codex App, and resolving it may require changes to the app's configuration, the system's resources, or the model's implementation.

Recommendation

Apply workaround: Consider reducing the context window size to a value closer to the effective limit of 258,000 tokens to avoid hitting the hardcoded cap, while awaiting further investigation or support from the Codex App team.