collaboration_modes ignores configured model + reasoning effort (hardcoded presets always override user config) #9783

What version of Codex is running?

0.89.0

What subscription do you have?

Pro

Which model were you using?

gpt-5.2, gpt-5.2-codex

What platform is your computer?

Microsoft Windows NT 10.0.19045.0 x64

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

Windows Terminal (WSL and PowerShell)

What issue are you seeing?

When the collaboration_modes feature is enabled, Codex doesn’t consistently respect user-selected model settings.
In particular, collaboration mode presets appear to override user config and CLI overrides for model and model_reasoning_effort.

What steps can reproduce the bug?

Steps to reproduce:

1. Enable collaboration modes (e.g. start Codex with --enable collaboration_modes).
2. Set model + reasoning effort via config and/or CLI, for example: - config.toml: model = "gpt-5.2-codex"
   model_reasoning_effort = "xhigh"
   • or CLI:
     • -m gpt-5.2-codex
     • -c 'model_reasoning_effort="xhigh"'
3. Start Codex and/or switch modes (e.g. Shift+Tab).

Actual behavior:

• The mode presets use hardcoded defaults (e.g. gpt-5.2-codex + medium) and completely ignores the configured values.

What is the expected behavior?

• Collaboration modes should respect the user’s chosen model and model_reasoning_effort (from config and/or CLI), and switching collaboration modes should not unexpectedly revert to hardcoded defaults.

Additional information

No response