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

# Models

> Choose which models Rubix uses for your organization and for one chat, and connect your own providers.

Rubix works straight away, with RubixKube's models included in your plan and no key to set up. Use **Settings > Models** when you have a reason to change that.

## When you'd use this

* **A better model for investigations.** Set a stronger Primary for chats and RCAs.
* **Lower cost on background work.** Set a smaller Worker for triage and titles.
* **Your data rules.** Connect your own provider so model traffic goes to your account. This needs Business or Enterprise.
* **One hard question.** Pick a different model for one chat without changing anything for anyone else.

![Settings, Models page with Primary and Worker models and the list of providers](https://mintcdn.com/rubixkube/uw89AXMFP3BXpNNP/images/console/settings-models-light.png?fit=max\&auto=format\&n=uw89AXMFP3BXpNNP\&q=85\&s=e56e97f36f61d9a6d7022b4525822a49)

## Set the default models

**Default models** has two roles. Each one has a model and a **Reasoning** level, for models that offer it.

| Role | Used for |
| - | - |
| **Primary** | Chats and RCAs. |
| **Worker** | Background work such as triage and titles. |

New chats start on Primary. Automations use Primary or Worker unless a step names a model.

If you leave Primary empty, it uses **RubixKube's default**. If you leave Worker empty, it's the **Same as Primary**.

Admins change these and click **Save**. Everyone else can see what's set.

## Pick a model for one chat

Click the model picker in the chat box, or type `/model`. Your pick stays with that chat and doesn't change anything for the organization. It follows the chat to any device.

If the model offers it, you can set **Reasoning** there too. **Auto** leaves it to the model.

## Connect your own provider

Bring your own model when you need a model we don't host, or want usage on your own account. This needs Business or Enterprise, and an admin.

These providers are available:

* Anthropic
* AWS Bedrock
* Google Gemini API
* Google Vertex AI
* OpenAI
* OpenRouter
* xAI
* **Your own endpoint**: any OpenAI-compatible server such as vLLM or LiteLLM, or your own Ollama

**1. Open the provider**

In **Settings > Models**, find it under **Available** and click to open it.

**2. Add your credentials and a model**

Fill in the key or the other fields the provider asks for, then pick a model.

**3. Click Test and connect**

Rubix tests the connection with that model. It's only saved if the test passes.

**4. Add more models**

Open the provider again and add any others. Each model has a switch that offers it in the chat.

Connected providers appear under **Enabled**, next to RubixKube. A provider that stops working shows **Needs attention** and the reason.

Removing a provider deletes its credentials and stops offering its models. Chats that used them switch to the default.

> [!NOTE]
>
> Want to keep your data off RubixKube's models? Turn on **Use only our own models** under **Own models only**. RubixKube's models stop being offered, and Primary and Worker run on your own models. Add a model of your own first.

## Related pages

> [card] [Chat with Rubix](https://docs.rubixkube.ai/using/rubix-chat-agent.md)
>
> Pick a model for one chat.

> [card] [Modes](https://docs.rubixkube.ai/using/modes.md)
>
> What Rubix can do without asking.

> [card] [Usage and plan](https://docs.rubixkube.ai/using/usage-and-plan.md)
>
> See what your plan includes.


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