Issue Station
Sign In
Tools hereToolbox

codex - 💡(How to fix) Fix Regression: GPT-5.4 / GPT-5.5 1M context setting is ignored in Codex IDE extension, stuck at 258k [2 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
openai/codex#20132•Fetched 2026-04-30 06:33:22
View on GitHub
Comments
2
Participants
2
Timeline
8
Reactions
0
Author
18120956431812095643
Participants
18120956431812095643github-actions[bot]github-actions[bot]
Timeline (top)
labeled ×4commented ×2closed ×1renamed ×1

Root Cause

This makes long coding sessions with GPT-5.4 / GPT-5.5 much less useful, because Codex starts auto-summarizing far earlier than expected. The main benefit of using the 1M context models is effectively unavailable inside the IDE extension.

RAW_BUFFERClick to expand / collapse

This appears to be a regression of the same context-window handling issue that happened when GPT-5.4 was first released. The model supports 1M context, but Codex IDE still behaves as if the usable context is capped at ~258k.

Impact:

This makes long coding sessions with GPT-5.4 / GPT-5.5 much less useful, because Codex starts auto-summarizing far earlier than expected. The main benefit of using the 1M context models is effectively unavailable inside the IDE extension.

What version of the IDE extension are you using?

latest

What subscription do you have?

Pro

Which IDE are you using?

VS Code

What platform is your computer?

Windows

What issue are you seeing?

The 5.4 and 5.5 models should support a 1M context window, but Codex appears to remain capped at around 258k context even after adding or changing the relevant 1M context configuration.

Because of this, Codex starts auto-summarizing very early, even when the actual conversation/workspace context is nowhere near 1M tokens. This makes long coding sessions difficult because useful context gets compressed or lost much earlier than expected.

This looks very similar to the issue that happened when 5.4 was first released: the model supported a larger context window, but Codex did not actually use it until the extension/config handling was fixed later. Now the same problem seems to be happening again with 5.5.

What steps can reproduce the bug?

Uploaded thread: 019dd732-5c39-77e3-9c28-d245c0bea2b4

What is the expected behavior?

Codex should respect the configured 1M context setting for models that support it, such as 5.4 and 5.5.

Configure Codex to use a 5.4 or 5.5 model that supports 1M context. Add or change the relevant configuration for 1M context. Start a long coding session in Codex. Observe that Codex still reports or behaves as if the usable context is around 258k. Auto-summarization begins much earlier than expected.

Additional information

The larger context window is one of the main reasons to use these models for coding tasks. If Codex is still capped at 258k internally, then the advertised 1M context support is effectively unusable inside Codex.

Please check whether the IDE extension / Codex runtime is correctly reading and applying the 1M context configuration for 5.4 and 5.5 models.

extent analysis

TL;DR

The issue can be addressed by verifying the correct configuration and application of the 1M context setting for GPT-5.4 and GPT-5.5 models in the Codex IDE extension.

Guidance

  • Check the Codex IDE extension configuration to ensure it is set to use the 1M context window for GPT-5.4 and GPT-5.5 models.
  • Verify that the extension is correctly reading and applying the 1M context configuration by checking for any updates or fixes related to context window handling.
  • Test the configuration by starting a long coding session and observing if Codex respects the 1M context setting and auto-summarizes at the expected point.
  • If the issue persists, review the uploaded thread (019dd732-5c39-77e3-9c28-d245c0bea2b4) for any additional clues or insights into the problem.

Notes

The issue seems to be related to a regression in the Codex IDE extension, similar to one that occurred when GPT-5.4 was first released. The fix may involve updating the extension or applying a specific configuration to enable the 1M context window for supported models.

Recommendation

Apply workaround: Check and update the Codex IDE extension configuration to ensure correct application of the 1M context setting for GPT-5.4 and GPT-5.5 models, as this is likely to resolve the issue and provide the expected behavior.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#container setup#orchestration issue#cache issue#memory leak#API versioning

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

codex - 💡(How to fix) Fix Regression: GPT-5.4 / GPT-5.5 1M context setting is ignored in Codex IDE extension, stuck at 258k [2 comments, 2 participants]