# Choosing a model

> Auto, or Vaulter Mini, Pro, and Max. What each model is for, what it costs in tokens, and when to switch.

Vaulter lets you choose which AI model powers your build, or pick for you. You will find the model picker near the chat in the build workspace.

## Auto and Manual

The picker has two settings:

- **Auto.** Vaulter picks the model for each step. This is the default, and the right choice for most people.
- **Manual.** You choose which model runs your requests, and it stays chosen until you change it.

## The models

In Manual mode there are three Vaulter agents, ordered from fastest to most capable:

- **Vaulter Mini.** Fast and cost-efficient. Great for small, well-defined changes: copy edits, styling tweaks, simple additions. Uses the fewest tokens.
- **Vaulter Pro.** Balanced performance. The all-rounder that handles the large majority of requests well. This is the default.
- **Vaulter Max.** Maximum intelligence. The deepest model, for genuinely complex, multi-step work: intricate features, tricky logic, and large changes that span many parts of the app. Available on paid plans.

On the free plan, Vaulter Max shows with a lock; upgrading unlocks it.

## Which to pick

- **Leave it on Auto** unless you have a reason not to.
- **Drop to Mini** when you are firing off lots of small changes and want the tightest loop at the lowest token cost.
- **Reach for Max** when a task feels genuinely complex, or when a request did not land the way you wanted and you want a more careful attempt before rephrasing.

## Models and tokens

Deeper models use more tokens per request than faster ones, and your plan includes a monthly token allowance. If you are watching your usage, Mini stretches your allowance furthest; Max spends it fastest but wastes fewer attempts on hard problems. See [Plans and pricing](/docs/account/plans).

One more place models matter: [Plan mode](/docs/getting-started/the-build-ide) always runs on the most capable model, regardless of your picker setting, because planning quality pays for itself.

## Frequently asked

#### What does Auto actually choose?

Vaulter matches the model to each step of the work, using deeper models where the request needs them and faster ones where it does not. It is tuned to be the right default, which is why Manual exists for preference, not necessity.

#### Does switching models affect my existing app?

No. The model only determines who handles your next request. Your app, its code, and its history are untouched by the picker.

#### Is Max worth the extra tokens?

For genuinely complex work, usually yes: one careful attempt is cheaper than three fast misses. For small clear changes it is overkill; that is what Mini is for.

#### Why is Vaulter Max locked for me?

Max is available on paid plans. On Free it shows with a lock, and upgrading unlocks it. See [Plans and pricing](/docs/account/plans).