# Choose the models Dvina uses

> Pick Dvina's default model, choose a model for each job, add your own provider, and use models that run on your computer.

Dvina chooses a model for you. If you'd rather choose yourself, Settings › **Model** lets you pick the **Default model**, and on Enterprise, choose a model for each job, add your own provider and use models that run on your own computers.

Choosing the model comes with the Max and Enterprise plans. On other plans, **Model** isn't in Settings. See [Plans and billing](/docs/help/plans-and-billing).

## Choose the default model

1. Open Settings and choose **Model**.
2. Next to **Default model**, choose a model from the list.

The list has Dvina's models and, on Enterprise, the models of a provider you added. The one Dvina uses now has a checkmark.

In a shared Livespace, its admins choose the Livespace's model. As a member, you choose between that model and the models of a provider you added yourself.

## Add your own provider

On Enterprise, you can connect a model provider of your own: one that gives you an API key and a base URL.

1. In Settings › **Model**, under **Custom Providers**, choose **Add Provider…**.
2. Next to **Provider**, choose the kind of provider you use. If yours isn't listed, choose the general option.
3. Enter the **Model**, the **Base URL** and the **API key** your provider gave you.
4. Click **Save**.

Dvina then uses that model, and it joins the **Default model** list. To change the provider later, choose it under **Custom Providers**. You only need to enter the **API key** again if you want to replace it.

To take a saved model off the list, open **Default model**, point to the model and click the delete button, then click **Delete**. It is no longer among the models you can choose.

> [!NOTE]
> The base URL must be a public HTTPS address, without a user name or password. If your Livespace's admins turned off members' own model providers, Dvina says so when you save.

## Choose a model for each job

On Enterprise, **Models by job** shows which model does each kind of work:

- **Main**
- **Quick tasks**
- **Images and scans**
- **Search index**
- **Transcription**
- **Voice conversations**
- **Image generation**

Each job shows its model, or **Dvina’s model** when none is chosen. Members can open a job to see its models; there, Dvina says **Its admins choose the models.**

If you're an admin:

1. Choose a job.
2. Choose **Add a model…** and pick one. **Main**, **Quick tasks** and **Images and scans** can have up to five models; the others take one.
3. Use **Move up** and **Move down** to set the order. The first model answers, and the next one takes over if it can't.
4. Click **Save**.

A model that can't do a job, or hasn't passed its capability test, can't be chosen for it, and Dvina says why. When you change the **Search index** model, Dvina reads everything in the Livespace's search index again in the background. Search keeps working with the current model until it's done.

### Set a data policy

Under **Data policy**, admins choose what may leave the company:

- **Standard:** A job without a chosen model is done by Dvina's.
- **No data out:** Only the company's own models are used. A job without one is off, and so are web search, connected apps and maps.
- **Region rule:** Only models in the region you enter are used, Dvina's included.

Here admins can also turn **Members’ own models** on or off, and **Turn off failing connections**.

### Connections

Under **Connections**, you see the model servers your Livespace uses. On Enterprise Cloud, Dvina manages them. When Dvina runs on your own servers, admins choose **Add a connection…**, enter its address and key, and choose **Run capability test** to see what the model can do. See [Where Dvina runs](/docs/teams/deployment).

## Use models on your computer

On Enterprise, Dvina can use models that run on your own Mac or PC. They're listed under **Models on Your Computer**.

1. Install Dvina for Mac or PC and sign in. See [Download and install Dvina](/docs/use-dvina/get-started/install).
2. Start a model server on that computer. Its models appear under the computer, with its memory and whether each model **Fits this computer**.
3. If you're an admin, click **Connect** next to a model. Dvina runs its capability test.
4. Under **Connected models**, choose **Use for…** and pick a job.

Turn on **Use Dvina’s model when this computer is off** so work goes on while the computer is off. With it off, chats say the model is offline, and background work waits for the computer.

To try a model again, choose **Test again**. To stop using it, choose **Remove**; give its jobs another model first.

> [!IMPORTANT]
> The model runs on your computer, but the conversation still passes through Dvina’s cloud, where the assistant works. While a model on your computer does a job, everyone in the Livespace whose work it does sends it to your computer.

### Get a model

When the computer can run models with Dvina itself, admins can choose **Get Models**. Under **Tested by Dvina**, you see models picked to fit this computer. Choose one, read its license, turn on **I accept the license**, then click **Install**. You can **Pause**, **Resume** or **Remove** it later.

### Allow only some models

Admins can keep a list under **Allowed Local Models**: which model servers members' computers may use, and under **Allowed Model Names**, which models. Only models whose name matches the list can be connected, and connected models that no longer match stop being used.

## Related

- [A tour of Settings](/docs/use-dvina/settings/overview)
- [Where Dvina runs](/docs/teams/deployment)
- [Plans and billing](/docs/help/plans-and-billing)
- [Dvina for teams and companies](/docs/teams/overview)