Skip to content

Expose relative cost of models in UI #20218

@alexandre-leng

Description

@alexandre-leng

What variant of Codex are you using?

codex app

What feature would you like to see?

Hi everyone,

Here’s a simple idea: not every request needs the most powerful model.

Right now, using AI can feel like driving a supercar just to run everyday errands. It works, but it’s often more than necessary, and it comes at a higher cost. For many basic tasks like quick questions, small edits, or simple reasoning, a lighter and more efficient model would be enough. Yet we often default to the most advanced option, which uses more tokens and shortens subscription value.

What could really improve the experience is giving users control over the level of intelligence they use, depending on the task.

Imagine being able to choose:

A lightweight model for simple tasks, using fewer tokens 0.25x TOKEN
A balanced model for more structured work 0.5x TOKEN
A full-capability model for complex problems 1x token

0.25x Token 0.5x Token and 1x token would appear in the interface to encourage developers and users to choose the right model and think about the level of complexity and intelligence they actually need. There’s no point in using a Ferrari to go grocery shopping. For simple tasks like translation, a much lighter model is more than enough.

Same credits, but used more efficiently.

This would allow users to rely on powerful models only when necessary, and use lighter ones the rest of the time. The result is clear: longer-lasting subscriptions, better cost control, and reduced compute usage.

At scale, this also has an environmental benefit, as it avoids unnecessary energy consumption.

More importantly, it introduces a different mindset. Instead of always using the most advanced model, users can focus on using the most appropriate one. The right level of intelligence for the right task.

This approach makes AI more practical, more sustainable, and better aligned with real needs.

In the end, most of us don’t need maximum power all the time. We just need the right amount, at the right moment.

Adding this kind of control directly into the interface would make a real difference.

Have a nice day
Alexander

Additional information

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    appIssues related to the Codex desktop appenhancementNew feature or requestrate-limitsIssues related to rate limits, quotas, and token usage reporting

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions