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

# Google AI

> Use your Google AI key to run Gemini Pro and Flash models in Rumus.

Bring your own Google AI Studio key to use Gemini in Rumus — including the long-context Pro models and the fast Flash variants.

## Before you start

You need:

* A Google account.
* An API key from [Google AI Studio](https://aistudio.google.com/apikey).
* Billing enabled on your Google AI project for paid models. Some Gemini models are available on a free tier with limits.

<Note>
  For **Vertex AI** (the Google Cloud enterprise endpoint), use the [OpenAI-compatible](/docs/models/openai-compatible) provider — Vertex requires service-account auth that's outside the scope of this provider type.
</Note>

## Add Google AI in Rumus

<Steps>
  <Step title="Open the model settings">
    Go to **Settings → AI → Models** and click **Add Model**.
  </Step>

  <Step title="Pick the provider">
    Set **Provider** to **Google AI**.
  </Step>

  <Step title="Paste your API key">
    Paste the key from Google AI Studio into **API Key**. It's stored encrypted in your local [vault](/docs/account/vault).
  </Step>

  <Step title="(Optional) Custom base URL">
    Leave **Base URL** blank to use `https://generativelanguage.googleapis.com`. Override only for special routing.
  </Step>

  <Step title="Pick a model">
    Choose from the list (Gemini 3 Pro, Gemini 3 Flash, Gemini 2.5 Pro/Flash, etc.) or toggle **Enter custom ID** to type a model ID manually.
  </Step>

  <Step title="Capabilities">
    On the **Capabilities** tab:

    * **Tool Calling** — Gemini supports function calling.
    * **Vision** — Gemini is natively multimodal.
    * **Prompt Cache** — supported on the Pro tier.
  </Step>

  <Step title="Save">
    The model appears in the picker under **Custom Models**.
  </Step>
</Steps>

## Recommended models

| Model                     | Good for                                                           |
| ------------------------- | ------------------------------------------------------------------ |
| **Gemini 3 Pro**          | Long-context analysis (whole repos, large logs); careful reasoning |
| **Gemini 3 Flash**        | Default daily driver — fast, low cost, strong tool use             |
| **Gemini 2.5 Flash-Lite** | High-volume autocomplete, classification                           |

Gemini Pro models accept extremely large contexts (up to 1M tokens on some versions), which is useful for reading entire repositories or long log files in a single pass.

## Tips

* **Thinking budget** is configurable per request on Pro models — Rumus exposes it as a per-thread setting next to temperature.
* **System instructions** are honored, but Gemini treats them slightly differently from OpenAI; use them for stable rules, put dynamic context in the user message.
* **Region matters** for latency. If you're far from `us-central1`, expect noticeable round-trip times on streaming responses.

## Troubleshooting

<AccordionGroup>
  <Accordion title="403 PERMISSION_DENIED">
    The API key is invalid or doesn't have the Generative Language API enabled. Re-create it in Google AI Studio.
  </Accordion>

  <Accordion title="429 RESOURCE_EXHAUSTED">
    You've hit the free-tier quota or a per-minute limit. Enable billing on the Google project or wait for the window to reset.
  </Accordion>

  <Accordion title="Model not in the list">
    Toggle **Enter custom ID** and paste the exact model ID (e.g. `gemini-3-pro-preview-12-01`).
  </Accordion>
</AccordionGroup>

<Note>
  Hit a snag we didn't cover? Ask in the [Rumus community](https://www.rumus.ai/community).
</Note>

## Next steps

<CardGroup cols={2}>
  <Card title="Other providers" icon="grid" href="/docs/models/overview">
    Anthropic, OpenAI, Z.AI, DeepSeek, Kimi, Ollama, OpenAI-compatible.
  </Card>

  <Card title="AI assistant" icon="robot" href="/docs/ai/overview">
    What you can do once a model is connected.
  </Card>
</CardGroup>