Add /smart mode to automatically choose model, effort, and planning mode

ChatGPT Feature requests
1

I’d like to suggest a /smart mode for Codex.

The idea is that when a user enters /smart, Codex automatically selects the best setup for the current task based on quality, speed, cost, task complexity, and risk.

Smart mode could automatically choose:

  • The best model for the task.
  • The right effort level.
  • Whether plan mode should be enabled.

For example:

  • Simple edits, formatting, small bug fixes, or basic commands → cheaper/faster model, low effort, no plan mode.
  • Medium tasks like adding a feature, fixing a localized bug, or modifying one part of the codebase → balanced model, medium effort, optional short plan.
  • Complex refactors, architecture changes, debugging across multiple files, migrations, or risky changes → stronger model, high effort, plan mode enabled.
  • Large or ambiguous tasks → plan mode first, then ask for confirmation before applying changes.

Possible UX:

/smart

Codex could show:

Smart mode enabled: automatically choosing model, effort, and planning behavior based on task complexity, cost, and risk.

Optional presets:

/smart economy
/smart balanced
/smart quality

Where:

/smart economy

prioritizes lower cost,

/smart balanced

balances cost and quality,

and:

/smart quality

prioritizes stronger reasoning and safer changes.

This would make Codex easier to use for people who care about cost, quality, and workflow safety but don’t want to manually manage model selection, effort level, and planning behavior for every task.

2

Thanks for the suggestion, @TitasK99.

Some of the pieces already exist today depending on the Codex experience. For example, some Codex integrations support automatic model selection, while Codex cloud tasks already handle planning internally. CLI and IDE users can also manually choose models, reasoning effort, and plan mode.

The main gap you're highlighting is having all of that bundled into a single /smart workflow that automatically adjusts model choice, effort level, and planning behavior based on task complexity and risk.

The /smart economy, /smart balanced, and /smart quality presets are a nice extension of that idea as well.

We'll make sure this gets forwarded to the internal team for logging and consideration.

-Mark G.