Tell Auto what to optimize for

Auto reads each message and picks a model for you. We covered how that choice happens in how Auto picks your model. Most of the time, that's the whole story — you type, Auto routes, you get an answer.

But there's a small preference you can set on top of it. It doesn't override the pick; it tells Auto which way to lean when the call is close — toward saving credits, staying balanced, or reaching for the best model your plan includes. Here's where the control is and when it's worth touching.

Where the control lives

After Auto answers, there's a small badge on the message showing which model it used. Open it. You'll see why Auto chose that model, and below that, a picker labeled Auto profile with four options:

• Auto — no explicit choice; use the default.
• Saver — favor lower-cost models to stretch your credits.
• Balanced — weigh credit spend, speed, and quality together.
• Best — go for the highest-quality result, spending credits faster.

Pick one and it takes effect from your next message. There's no separate settings page to hunt for — the preference sits right next to the answer it affects.

Leave it on the default for everyday work

If you've never touched this, Auto runs on Balanced, and that's the right place to leave it for most work.

Balanced doesn't mean lowest-common-denominator. Auto still does everything from the previous post: it reads the message, escalates when it's unsure, thinks longer on genuinely hard problems, and can take a second look when a first answer is clearly weak. Balanced just means it also keeps an eye on credit spend and speed instead of always reaching for the most expensive model.

For the vast majority of messages, you won't get a better answer by changing this. Leave it alone.

Drop to Saver for quick, throwaway questions

Saver leans toward lower-cost models — which are usually the faster ones too.

It's a good fit when you don't need the strongest model in the room:

• A quick fact you'll verify anyway.
• Rephrasing a sentence or fixing tone.
• Casual back-and-forth where speed matters more than depth.
• Any question you'd forget five minutes later.

The point is to stretch your monthly credit further on the messages that don't need much. Auto still reads each message, so a genuinely hard one can still pull a stronger model — Saver lowers the everyday floor, it doesn't cap what Auto is allowed to do.

Move to Best when the answer really matters

Best goes for the highest-quality model your plan includes, and it spends credits faster to get there.

Reach for it when the output is worth the credits:

• Writing you're about to send.
• Analysis where a wrong conclusion costs you.
• Code you'll actually ship.

Best is available on paid plans. On the free plan the picker isn't adjustable — free accounts route through a set of fast everyday models, and Best in particular needs a paid plan. What each plan includes is current at the time of writing; see pricing for live details.

It's a standing preference, not a per-message switch

Whatever you pick applies to every new message until you change it again, and it's saved to your account — so it follows you across devices, not just this one chat.

That's worth remembering. If you switch to Saver for a batch of quick questions and forget about it, later messages stay on Saver too. To go back to the default, open the picker and choose Auto — it clears your choice and returns you to Balanced.

In short

Leave it on the default for everyday work — Balanced already leans on quality where it counts. Drop to Saver for quick throwaway questions and to make your credits last. Move to Best when you want the strongest model on every message and don't mind spending credits faster to get it.