inference.eu

Chat · OpenGPT-X (Fraunhofer, Forschungszentrum Jülich, TU Dresden, DFKI)

Run Teuken 7B in the EU through an OpenAI-compatible API

Teuken 7B Instruct is a compact multilingual model from the German research project OpenGPT-X, trained on all 24 official EU languages. The commercial version is released under Apache 2.0. On inference.eu you call it with the OpenAI or Anthropic request format, with requests processed in EU data centres and never used for training.

Contact salesExample request
  1. 017B
  2. 02EU data centres
  3. 03Apache 2.0

What Teuken 7B is

Teuken 7B was developed in the OpenGPT-X project by Fraunhofer, Forschungszentrum Jülich, TU Dresden and DFKI, with funding from the German Federal Ministry for Economic Affairs. It is a 7-billion-parameter instruction-tuned model, pre-trained on about 4 trillion tokens covering all 24 official languages of the European Union.

On inference.eu we serve Teuken-7B-instruct-commercial-v0.4, the version released under Apache 2.0 for commercial and research use. A later version, v0.6, was trained on more data but is licensed for non-commercial use only. Teuken's instruction tuning used English and German datasets plus instructions in 22 other European languages, and its context window is 4,096 tokens.

What it does well, and where it falls short

Teuken's point is balance across European languages. Because it was not built English-first, it gives more even results across the 24 EU languages than many models of its size. Its small size makes it inexpensive to run, which suits short, high-volume tasks such as classifying messages, rewriting short texts or answering simple questions in many languages.

It has clear limits. The 4K context means it cannot read long documents in one go, it does not process images, and on reasoning, code and complex instructions it is well behind larger and newer models. The model card also warns that it may produce biased or inappropriate output. For harder tasks, EuroLLM 22B or Mistral Small 4 are better choices.

Calling Teuken 7B through the API

inference.eu uses the OpenAI request format, so the OpenAI SDKs, LangChain, LlamaIndex and similar tools work without changes. Set the base URL to the inference.eu API, use your inference.eu key and set the model to teuken-7b-instruct-commercial.

Teuken's chat template has a language setting that tells the model which language to use for its built-in system prompt. Keep your prompts short, because system prompt, conversation history and answer all have to fit in 4,096 tokens. Streaming works as usual, and the Anthropic Messages format is also available if your tools were written for Claude.

A German model, processed in the EU

Teuken was built with public German funding on European research infrastructure, and on inference.eu it runs in Europe too. Your prompts and its answers are processed in data centres in the European Union, and they are never used to train any model.

Every customer gets a data processing agreement under Article 28 of the GDPR together with our subprocessor list, so you can record exactly who handles the data. Billing is in euros. The commercial v0.4 release is under Apache 2.0, which allows commercial use without user limits. Do not confuse it with v0.6, whose CC BY-NC 4.0 licence rules out commercial use.

teuken-7b-instruct-commercial · 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="teuken-7b-instruct-commercial",
    messages=[{"role": "user", "content": "Summarise this contract in five bullet points."}],
)
print(response.choices[0].message.content)

What teams use it for

  • 1

    Short multilingual replies

    Generate short answers, confirmations and rewrites in any of the 24 EU languages for forms, chat widgets and notifications.

  • 2

    Message classification

    Sort incoming messages by language, topic or urgency where a small, inexpensive model is enough.

  • 3

    Research and public projects

    Use a model developed by German public research institutions for projects that want a European foundation.

Teuken 7B Instruct: common questions

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

Ask a question

Related

All models

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