> ## 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.

# DeepMask D1 & D2

> DeepMask's own routing models: D1 picks the best model for each prompt, and D2 does the same using only models hosted in Germany.

DeepMask D1 and D2 aren't single models. They are DeepMask's own routers: when you send a message, they score how complex it is and pass it to a suitable model from a pool. Simple questions go to fast, efficient models; demanding ones go to more capable models. You get a good answer without choosing a model yourself.

> [card] DeepMask D1
>
> Routes each prompt to the best model pool based on its complexity.

> [card] DeepMask D2
>
> Uses the same routing, but only picks from models hosted in Germany. Use it when your data must stay on German infrastructure.

## When to use them

* **You're not sure which model to pick.** Start with D1 and let DeepMask decide per message.
* **Your conversations mix simple and hard questions.** The router matches each one to an appropriate model instead of using one model for everything.
* **You need German hosting without comparing models.** D2 keeps every message on German-hosted models.

> [!TIP]
>
> When you know exactly what a task needs — for example image analysis or a very long document — choose a specific model instead. See [Choosing a model](https://documentation.deepmask.io/models/choosing-a-model.md).

## Specifications

| Specification | DeepMask D1 | DeepMask D2 |
| - | - | - |
| Provider | DeepMask | DeepMask |
| Hosting | EU | Germany |
| Context window | 111K tokens | 200K tokens |
| Max output | 64K tokens | 131K tokens |
| Image input | No | No |
| Tool use | Yes | Yes |

[Try DeepMask D1 and D2 →](https://chat.deepmask.io/)


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