›_ Codex Issues
Codex Issues / #12495

/review can submit stale reasoning.effort after switching to gpt-5.1-codex-mini

Open 💬 2 comments Opened Feb 22, 2026 by GraciousGazelles
💡 Likely answer: A maintainer (hdcodedev, contributor) responded on this thread — see the highlighted reply below.

What version of Codex CLI is running?

Source build from openai/codex commit 66d5d34e6 (binary reports codex-cli 0.0.0).

What subscription do you have?

Pro

Which model were you using?

gpt-5.3-codex-spark (xhigh) -> gpt-5.1-codex-mini (high)

What platform is your computer?

Linux x86_64

What terminal emulator and version are you using (if applicable)?

tmux 3.5a

What issue are you seeing?

/review can send stale reasoning.effort = xhigh after switching to gpt-5.1-codex-mini, causing:

Unsupported value: "xhigh" is not supported with the "gpt-5.1-codex-mini" model. Supported values are: "low", "medium", and "high".

What steps can reproduce the bug?

  1. Start a session on gpt-5.3-codex-spark with effort xhigh.
  2. Switch to gpt-5.1-codex-mini with effort high.
  3. Run /review.
  4. Observe unsupported_value on reasoning.effort.

What is the expected behavior?

/review should use current runtime model/effort (or clamp to supported effort for the selected model).

Additional information

None.

View original on GitHub ↗

2 Comments

hdcodedev contributor · 6 months ago

This issue has been successfully reproduced. I am performing final checks and testing.
Once everything is complete, I will move PR https://github.com/openai/codex/pull/12630 to review.

miaopan607 · 3 months ago

I reproduced this locally.

/review appears to read stale config.model_reasoning_effort instead of the current thread’s effective runtime setting. So after changing effort via /model, normal turns use the new value but /review can still submit the old one.

A likely fix is to make the review child turn inherit from the parent turn’s effective config / turn context, then apply review-only overrides on top.