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

# Gemma

> Google's open-weight Gemma models in DeepMask, hosted in Germany: Gemma 4 31B with tool use, and Gemma 3 27B on StackIT, both with image input.

Gemma is Google's family of open-weight models. In DeepMask, both Gemma models run in Germany, so your data stays on German infrastructure. Both accept images and documents and have a 131K-token context window.

> [tab] Gemma 4 31B
>
> ## About
>
> Gemma 4 31B is the newer and larger Gemma model. Unlike Gemma 3 27B, it supports tool use, so it can work with connectors, web search, and skills. Choose it when you need a German-hosted model that can also take action through tools.
>
> ## Specifications
>
> | Specification | Value |
> | - | - |
> | Model provider | Google |
> | Hosting | Germany |
> | Context window | 131K tokens |
> | Max output | 131K tokens |
> | Image input | Yes |
> | Tool use | Yes |
> | Reasoning | Standard |

> [tab] Gemma 3 27B (StackIT)
>
> ## About
>
> Gemma 3 27B runs on StackIT, the German sovereign cloud operated by the Schwarz Group. It reads text and images and answers in text. It doesn't support tool use, so use it for chat, writing, and document or image analysis rather than connector-based tasks.
>
> ## Specifications
>
> | Specification | Value |
> | - | - |
> | Model provider | Google |
> | Hosting | Germany (StackIT) |
> | Context window | 131K tokens |
> | Max output | 131K tokens |
> | Image input | Yes |
> | Tool use | No |
> | Reasoning | Standard |

[Try Gemma in DeepMask →](https://chat.deepmask.io/)


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