> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rafftechnologies.com/llms.txt
> Use this file to discover all available pages before exploring further.

# App sizes, scaling and sleeping

> Raff Apps instance sizes, fixed instances, autoscaling on requests, scale to zero and the instance limits.

<sub>Updated October 19, 2026</sub>

Every instance of an app has the same size: a share of CPU, an amount of memory and a temporary disk. You choose how many instances run, whether Raff adds instances under load, and whether a web or private app sleeps when nobody uses it.

## How it works

### Sizes

| Size | vCPU | Memory | Temporary disk | Good for |
| - | - | - | - | - |
| Micro | 0.25 | 512 MB | 2 GB | Tests and hobby projects |
| Starter | 0.5 | 1 GB | 5 GB | Small sites and APIs |
| Standard | 1 | 2 GB | 10 GB | Most production apps |
| Pro | 2 | 4 GB | 20 GB | Busy apps |
| Pro Max | 4 | 8 GB | 40 GB | Heavy work |
| Scale | 8 | 16 GB | 80 GB | The largest apps |

The create page lists sizes by their specs, for example **0.5 vCPU · 1 GB**. For prices, see [Raff Apps pricing](/products/build/raff-apps/details/pricing).

An app cannot use more CPU than its size: a busy app runs slower instead. An app that runs out of memory is stopped and restarted. Change the size in **Settings**, **Instance size**; the app restarts on the new size.

### Fixed instances

Web, private and worker apps run 1 to 6 instances. Each instance runs all the time and is billed at the size's price. Set the number in **Settings**, **Scaling**, **Instances**.

### Autoscaling

Web apps, and private apps that scale to zero, can scale between **Min instances** and **Max instances**. Raff adds instances as the number of requests in progress per instance grows, and removes them as traffic drops. Workers and private apps that run all the time use a fixed number of instances.

### Scale to zero

Web and private apps can sleep. With **Scale to zero** on, the app sleeps about 15 minutes after its last request and wakes on the next one. The first request after sleeping takes a few seconds longer while an instance starts.

On pay-as-you-go, a sleeping app costs nothing; the minutes it is awake are billed like any running time. With autoscaling, scale to zero lets **Min instances** be 0.

### Instance limits

| Limit | Value | Notes |
| - | - | - |
| Instances per app | 6 | Fixed instances, or the autoscaling maximum |
| Instances per account | 20 | Web, private and worker apps. Paused, failed and not yet deployed apps count too, because resuming starts them again |

Cron and job apps do not count. Scaling down is always allowed, even above a limit. Contact support to raise either limit.

## When to use it

* **One fixed instance**: most small apps, and any app that keeps state in memory.
* **Several fixed instances**: steady traffic, no waiting for new instances.
* **Autoscaling**: traffic that rises and falls during the day.
* **Scale to zero**: test apps, internal tools and sites with little traffic.

## Trade-offs

* The first request after sleeping is slower by a few seconds.
* Workers cannot sleep or autoscale.
* Instances do not share memory or disk. Keep shared data in a [database](/products/build/raff-apps/concepts/database-links).
* On monthly billing, the size cannot be changed from the dashboard yet. Contact support to change it.

## Related

<CardGroup cols={2}>
  <Card title="Scale an app" icon="rocket" href="/products/build/raff-apps/quickstart-guides/scale-an-app">
    Instances, autoscaling and scale to zero.
  </Card>

  <Card title="Pricing" icon="circle-dollar" href="/products/build/raff-apps/details/pricing">
    The price of every size.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.