Plan mode overrides explicitly selected high/xhigh effort to medium

What version of Codex CLI is running?

codex-cli 0.101.0

What subscription do you have?

ChatGPT Pro (Sign in with ChatGPT)

Which model were you using?

gpt-5.3-codex-high (also reproducible with gpt-5.3-codex-xhigh)

What platform is your computer?

Darwin 25.2.0 arm64 arm

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

Ghostty (TERM=xterm-ghostty), tmux 3.6a

What issue are you seeing?

I selected a high-effort model (gpt-5.3-codex-high / gpt-5.3-codex-xhigh), and /model continues to show that selection.

However, in Plan mode, behavior and/or UI indication appears to fall back to medium effort.
So there seems to be a mismatch between what /model reports and what Plan mode effectively uses/displays.

What steps can reproduce the bug?

1. Start a new Codex CLI session.
2. Select gpt-5.3-codex-high or gpt-5.3-codex-xhigh.
3. Run /model and confirm high/xhigh is selected.
4. Enter Plan mode using /plan or Shift+Tab.
5. In Plan mode, observe effort indicator/behavior appears medium.
6. Run /model again and note it may still report high/xhigh.
7. This shows a Plan-mode effective effort vs /model reported effort mismatch.

What is the expected behavior?

Plan mode and /model should agree on the effective effort.

If high/xhigh was explicitly selected, Plan mode should keep that selection.
If Plan mode intentionally overrides effort, /model should clearly show the effective override.

Additional information

I understand external code contributions may not be accepted.
I am sharing a detailed report with repro steps, environment, root-cause hypothesis, and a possible fix direction.

Environment:

• Codex CLI: 0.101.0
• OS: macOS 26.2 (Build 25C56)
• Platform: Darwin 25.2.0 arm64 arm
• Terminal: Ghostty (TERM=xterm-ghostty)
• Multiplexer: tmux 3.6a

Root-cause hypothesis:

• Plan preset default effort is medium.
• Plan-mode mask/effective-mode logic may differ from what /model reports.
• Result: reported selected effort and effective Plan-mode effort can diverge.

Possible fix direction:

1. Define a single source of truth for "effective effort" per active mode.
2. Make /model show effective values (or both selected vs effective with clear labels).
3. Preserve explicitly selected high/xhigh when entering Plan mode unless user opts into override.