codex app does not respect model_reasoning_effort

What version of the Codex App are you using (From “About Codex” dialog)?

260203.1501 (523)

What subscription do you have?

plus

What issue are you seeing?

in .codex/config.toml

model = "gpt-5.2"
model_reasoning_effort = "high"

When I start codex.app, the reasoning level of a new thread is High,
but when click a previous thread, the reasoning level becomes medium.

What steps can reproduce the bug?

1. Create a new thread in the Codex tui. Right now it uses GPT-5.2 with model_reasoning_effort=high.
2. Select this session in the Codex app. The UI shows GPT-5.2 medium.

What is the expected behavior?

_No response_

Additional information

_No response_