> For the complete documentation index, see [llms.txt](https://langtech-bsc.gitbook.io/applied-intelligence/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://langtech-bsc.gitbook.io/applied-intelligence/models/machine-translation.md).

# Machine translation

SalamandraTA: translation LLMs covering the languages of Spain and Europe, with terminology, formatting and document-level support.

Our translation models belong to the **SalamandraTA** family: Salamandra LLMs that were further pre-trained on parallel data and then instruction-tuned for translation tasks. They translate between dozens of languages in any direction, with a particular strength in **Catalan, Spanish, Basque, Galician, Asturian, Aragonese and Aranese**.

## Recommended models

| Model                                                                                                                                       | Size | Languages        | License    | Best for                                                               |
| ------------------------------------------------------------------------------------------------------------------------------------------- | ---- | ---------------- | ---------- | ---------------------------------------------------------------------- |
| [**salamandraTA-7b-instruct**](https://langtech-bsc.gitbook.io/applied-intelligence/models/machine-translation/salamandrata-7b-instruct.md) | 7.8B | 40 + 3 varieties | GPL-3.0    | Best quality. Terminology, HTML/XML, register, post-editing            |
| [**salamandraTA-2b-instruct**](https://langtech-bsc.gitbook.io/applied-intelligence/models/machine-translation/salamandrata-2b-instruct.md) | 2.3B | 35 + 3 varieties | Apache 2.0 | Fast, cheap translation on modest hardware, under a permissive license |

Both come in full precision (Transformers / vLLM) and **GGUF** builds for llama.cpp:

| Build              | File size | Runs on                                                     |
| ------------------ | --------- | ----------------------------------------------------------- |
| 7B · BF16          | ≈ 15.5 GB | A single 24 GB GPU                                          |
| 7B · GGUF Q4\_K\_M | 4.85 GB   | A laptop or small GPU (about 1–2 BLEU below full precision) |
| 2B · GGUF Q8\_0    | 2.40 GB   | CPU or any GPU                                              |
| 2B · GGUF Q4       | 1.52 GB   | CPU, edge devices                                           |

{% hint style="success" %}
**Replacing a commercial MT API?** BSC's [mt-wrapper](https://github.com/langtech-bsc/mt-wrapper) exposes SalamandraTA behind Google Translate and DeepL-compatible endpoints, so existing integrations can switch with minimal changes.
{% endhint %}

## Prompting basics

SalamandraTA models are **not chat models**. They expect a task prompt in the user turn and need no system prompt. The basic translation prompt is:

```
Translate the following text from {source} into {target}.
{source}: {text}
{target}:
```

Write language names in full, in English (`Spanish`, `Catalan`, `Basque`…), as in the model cards' examples. Each model page lists the prompts for its other tasks.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://langtech-bsc.gitbook.io/applied-intelligence/models/machine-translation.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
