GPT models consistently run out of context earlier than indicated

[iloveicedgreentea]

Description

With 5.1, 5.2, and 5.2, I get context length errors at 67% every time and have to do manual compaction to fix it

Plugins

No response

OpenCode version

No response

Steps to reproduce

No response

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• GPT 5.2 Codex - Compaction happens way too early and context limit is wrong #11086: GPT 5.2 Codex - Compaction happens way too early and context limit is wrong
• [BUG]: GPT 5.2 context_length_exceeded / Your input exceeds the context window of this model. #6071: [BUG]: GPT 5.2 context_length_exceeded / Your input exceeds the context window of this model.
• Request size exceeds model context window #10220: Request size exceeds model context window
• Auto-compaction enabled by default, but context_length_exceeded errors still occur in agent workflows #8089: Auto-compaction enabled by default, but context_length_exceeded errors still occur in agent workflows
• Compaction overflow check doesn't account for large tool outputs until the next step, causing context overflow #10634: Compaction overflow check doesn't account for large tool outputs until the next step, causing context overflow

Feel free to ignore if none of these address your specific case.

[jesussmile]

same issue here

[jasgeo75]

Same, broken support for both the gpt-5.x-codex and opus 4.6 models is seriously impeding workflow.

[heswithme]

yup, gotta compact manually around 200k tokens, otherwise session gets borked

[entrptaher]

Facing same situation on gpt codex 5.3, with where it says,

271,647  68% ($0.00) v1.1.5

It was already suspecious that it allows 270k tokens with 68% usage mark. This confirms the same finding with the OP as well.

Plus, the /compact didn't actually compact it and allow me to continue, rather it asked me to copy a super stripped down version of the discussion and paste it in a new session.