Add mode-specific default chat model and reasoning effort preferences
What variant of Codex are you using?
App and IDE Extension
What feature would you like to see?
Title: Add mode-specific default chat model and reasoning effort preferences
Priority: High
Why High Priority:
This can significantly optimize user-side cost by allowing cheaper/faster defaults for routine chat while reserving higher reasoning effort only for Plan mode or Pursue Goal mode. It also reduces repeated manual switching and helps users avoid accidentally using expensive model or reasoning settings for low-complexity work.
Summary
Add user-configurable defaults for chat model and reasoning effort per Codex mode:
- Standard chat
- Plan mode
- Pursue Goal mode
These defaults should apply both when creating a new thread and when switching modes inside an existing thread.
Problem
A single global model and reasoning-effort default is too coarse. Different workflows have different cost and quality needs:
- Standard chat usually benefits from fast, cost-efficient defaults.
- Plan mode may benefit from stronger reasoning for design and tradeoff analysis.
- Pursue Goal mode may need higher reasoning for longer-running implementation work.
Without mode-specific defaults, users must repeatedly adjust settings manually and may unintentionally spend more than needed.
Proposed Behavior
Add preferences for:
- Standard chat default model
- Standard chat default reasoning effort
- Plan mode default model
- Plan mode default reasoning effort
- Pursue Goal mode default model
- Pursue Goal mode default reasoning effort
When a user starts a new thread, Codex should apply the defaults for the selected mode.
When a user switches modes inside an existing thread, Codex should apply the defaults for the new mode if the override flag is enabled.
If the configured model is unavailable, Codex should automatically fall back to the built-in Codex default model for that mode.
Override Behavior
Add or support an override preference flag:
- If enabled, switching modes replaces the current model and reasoning effort with that mode’s configured defaults.
- If disabled, switching modes preserves the current thread/session model and reasoning effort.
Default/Fallback Behavior
- Existing users should keep the current Codex default behavior unless they explicitly configure mode-specific preferences.
- If a user has not configured a preference for a mode, use the default Codex model and reasoning effort for that mode.
- If a configured model becomes unavailable, fall back automatically to the default Codex model for the active mode.
Acceptance Criteria
- Users can configure model and reasoning effort separately for Standard, Plan, and Pursue Goal modes.
- Defaults apply when creating a new thread.
- Defaults apply when switching modes if the override flag is enabled.
- Current-session settings are preserved on mode switch if the override flag is disabled.
- Missing or unavailable configured models automatically fall back to Codex defaults.
- Preferences persist across app restarts.
- UI clearly distinguishes persistent mode defaults from temporary per-thread settings.
- Existing users see no behavior change unless they opt into custom preferences.
Additional information
_No response_
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗