---
title: "EuroLLM 22B API hosted in the EU"
description: "Run EuroLLM 22B Instruct, the European open model trained for all 24 official EU languages, on inference.eu with an OpenAI-compatible API."
canonical: "https://inference.eu/models/eurollm-22b-instruct/"
language: "en"
---

Chat · UTTER project (Instituto Superior Técnico, Unbabel, University of Edinburgh and partners)

# Run EuroLLM 22B in the EU through an OpenAI-compatible API

EuroLLM 22B Instruct is a European open model built to understand and write every official language of the European Union. It was trained on the MareNostrum 5 supercomputer through a EuroHPC grant. On inference.eu you call it with the OpenAI or Anthropic request format, processed in EU data centres and never used for training.

[Contact sales](https://inference.eu/contact/)[Example request](#example)

1.  0122.6B
2.  02EU data centres
3.  03Apache 2.0

## What EuroLLM 22B is

EuroLLM is a family of open language models developed by a consortium of European universities and companies, led from Instituto Superior Técnico in Lisbon and including Unbabel, the University of Edinburgh, Université Paris-Saclay, Sorbonne Université and the University of Amsterdam. The 22B Instruct model, version 2512, has about 22.6 billion parameters and a context window of 32,768 tokens.

It was trained on around 4 trillion tokens using 400 H100 GPUs on MareNostrum 5 in Barcelona. The training mix was deliberately spread across all 24 official EU languages plus Arabic, Catalan, Chinese, Galician, Hindi, Japanese, Korean, Norwegian, Russian, Turkish and Ukrainian, with parallel data to strengthen translation.

## What it does well, and where it falls short

EuroLLM's strength is breadth across European languages. Where many models are strong in English and a handful of large languages, EuroLLM was built to produce usable text in Estonian, Latvian, Maltese, Irish, Slovenian or Croatian as well. That makes it a good fit for translation, multilingual customer communication and public-sector services that must treat every EU language equally.

It is a dense 22B model with a 32K context, so it is not built for very long documents, image input or complex agentic coding. On reasoning-heavy tasks and code, larger models such as Qwen3 or Mistral Small 4 do better. Use EuroLLM where language coverage matters more than raw reasoning power.

## Calling EuroLLM 22B through the API

inference.eu uses the OpenAI request format, so the OpenAI SDKs for Python and JavaScript, LangChain, LlamaIndex and other compatible tools work as they are. Set the base URL to the inference.eu API, use your inference.eu key and set the model to `eurollm-22b-instruct`.

Write your system prompt and user messages in the language you want back; the model follows the conversation language. For translation, name the source and target language explicitly in the prompt. Streaming works as usual, and the Anthropic Messages format is also available if your tools were built for Claude.

## A European model, processed in the EU

EuroLLM was funded and trained in Europe, and on inference.eu it also runs in Europe. Your prompts and its responses are processed in data centres in the European Union, and they are never used to train any model.

Every customer receives a data processing agreement under Article 28 of the GDPR and our list of subprocessors, so you can document the full chain of processing. Billing is in euros. The Apache 2.0 licence allows commercial use without user or revenue limits, which suits public administrations and companies that want a European model with no strings attached.

## EuroLLM 22B Instruct

Maker

UTTER project (Instituto Superior Técnico, Unbabel, University of Edinburgh and partners)

Made in

Portugal

Type

Chat

Parameters

22.6B

Context window

32,768 tokens

Licence

[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)

Released

December 2025

Model ID

eurollm-22b-instruct

Processing

EU data centres

Training on your data

Never

Source

[huggingface.co](https://huggingface.co/utter-project/EuroLLM-22B-Instruct-2512)

eurollm-22b-instruct · Python

```
import os
from openai import OpenAI

client = OpenAI(base_url="https://api.inference.eu/v1", api_key=os.environ["INFERENCE_EU_API_KEY"])

response = client.chat.completions.create(
    model="eurollm-22b-instruct",
    messages=[{"role": "user", "content": "Summarise this contract in five bullet points."}],
)
print(response.choices[0].message.content)
```

## What teams use it for

-   1

    ### Translation across EU languages

    Translate product texts, notices and correspondence between all 24 official EU languages, including smaller ones such as Maltese and Irish.

-   2

    ### Multilingual citizen services

    Answer questions from citizens or customers in their own language, with the same assistant for every member state.

-   3

    ### Localising content

    Adapt marketing copy, help-centre articles and documentation for each European market from one source text.

## EuroLLM 22B Instruct: common questions

The questions buyers, developers and legal teams ask us most. Something missing? Ask us directly.

GDPR compliance depends on how you process data. On inference.eu, EuroLLM requests are processed in EU data centres, not used for training and covered by a data processing agreement under Article 28 GDPR with a published subprocessor list. That gives you a contractual basis to process personal data with it.

All 24 official EU languages: Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Irish, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish and Swedish. It also covers Arabic, Catalan, Chinese, Galician, Hindi, Japanese, Korean, Norwegian, Russian, Turkish and Ukrainian.

EuroLLM 22B handles 32,768 tokens of context. That covers long emails, articles and reports of a few dozen pages. For very long documents, split them into sections or use a model with a larger window, such as Mistral Small 4 or Apertus.

Yes. EuroLLM 22B Instruct is released under the Apache 2.0 licence, which allows commercial use, modification and redistribution without user limits or fees. You do not need an agreement with the consortium that built it.

EuroLLM was developed by Instituto Superior Técnico and Instituto de Telecomunicações in Lisbon, the University of Edinburgh, Aveni, Unbabel, Université Paris-Saclay, Artefact Research Center, the University of Amsterdam, Naver Labs and Sorbonne Université, with compute from EuroHPC.

Both target all 24 EU languages. EuroLLM 22B is about three times larger, has a longer context of 32K tokens and adds eleven non-EU languages. Teuken 7B is smaller and cheaper to run, with a 4K context. For most production use, EuroLLM 22B is the stronger choice.

It can help with simple code and logic, but it was not built for that. Its focus is language coverage and translation. For programming, use Devstral Small 2 or Qwen3.6, and for multi-step reasoning a model such as Magistral Small or DeepSeek V4 Flash.

Point the base URL of your OpenAI client at the inference.eu API, use your inference.eu key and set the model to \`eurollm-22b-instruct\`. Your message format and streaming code stay the same.

Yes. The consortium also publishes EuroLLM 9B Instruct and earlier 1.7B models with the same language coverage. The 22B model is the most capable of the family and the one to start with for quality-sensitive tasks.

No. EuroLLM 22B is a text-only model. For images, scanned documents or screenshots, use a multimodal model such as Mistral Small 4 or Gemma 4, or DeepSeek-OCR 2 to turn documents into text first.

[Ask a question](https://inference.eu/contact/)

## Related

[All models](https://inference.eu/models/#directory)

-   [Teuken 7B InstructRun Teuken 7B Instruct, the German open model trained on all 24 official EU languages, on inference.eu through an OpenAI-compatible API.Chat](https://inference.eu/models/teuken-7b-instruct/)
-   [Apertus 1.5 70BRun Apertus 1.5 70B, the fully open Swiss model from EPFL and ETH Zurich, on inference.eu with an OpenAI-compatible API and EU data processing.Chat](https://inference.eu/models/apertus-70b/)
-   [Mistral Small 4 119BRun Mistral Small 4 119B from French lab Mistral AI on inference.eu: reasoning, vision and code in one model, processed in EU data centres.Chat](https://inference.eu/models/mistral-small-4-119b/)
-   [Llama 3.3 70B InstructRun Meta's Llama 3.3 70B Instruct on inference.eu: an OpenAI-compatible API with requests processed in EU data centres and billed in euros.Chat](https://inference.eu/models/llama-3-3-70b-instruct/)

## Move your AI workloads to EU infrastructure

Tell us which models you use and roughly how much traffic you have. We reply within one working day with pricing and next steps.

-   Reply within one working day
-   Pricing for your volume
-   DPA ready to sign