The European research project OpenGPT-X recently published a large language model called “Teuken-7B”. The model comprises seven billion parameters and was trained from scratch with the 24 official languages of the EU, including German. The AI is free of charge and can be used by companies without sensitive data remaining outside the company.
“Teuken-7B” is an open source model developed by a consortium led by the Fraunhofer Institutes IAIS and IIS. The model was trained using the JUWELS supercomputer at Forschungszentrum Jülich and contains around 50% non-English pre-training data.
Deutsche Telekom has joined forces with OpenGPT-X to create an offering for AI models “Made in Germany”. The Teuken-7B models are available to users of generative AI for immediate use and can be used as a powerful, trustworthy and open-source alternative to the major conventional models.
The OpenGPT-X project will run until March 31, 2025 and offers partners the basis for training further proprietary models.
Further information on this topic can be found here: