Codex CLI resets `model_reasoning_effort` to Medium (with proof)

What version of Codex CLI is running?

codex-cli 0.102.0

What subscription do you have?

plus

Which model were you using?

gpt-5.3-codex

What platform is your computer?

Linux 6.6.87.2-microsoft-standard-WSL2 x86_64 x86_64

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

Windows Terminal (WSL)

What issue are you seeing?

When I manually select a reasoning level for gpt-5.3-codex (or any other model)—for example High, Low, or Extra High—the CLI switches back to Medium.

What steps can reproduce the bug?

To reproduce the issue:

1. Select gpt-5.3-codex and set the Reasoning Level to High or Extra High.
2. Type "hi".
3. After the model replies, press Shift + Tab to switch to Plan Mode, then type "hi" again.
4. You’ll see the bug: the CLI switches back to gpt-5.3-codex (Medium).

Additional context: the config file does not change—it stays set to the reasoning level I selected. The switch back to Medium only appears in the TUI, and I verified this.

<img width="408" height="328" alt="Image" src="https://github.com/user-attachments/assets/449fb56b-1fa0-4f2e-affa-158f371fec5e" />

and this is the config file:

<img width="275" height="62" alt="Image" src="https://github.com/user-attachments/assets/91097e81-844a-4f03-9a85-4ada1cd1d3be" />

I hope this makes it clear now @etraut-openai

What is the expected behavior?

_No response_

Additional information

_No response_