Europe currently has few open language models that are large enough to serve as a basis for industrial applications. Reasoning models that can solve complex tasks are also hardly available from European sources, meaning that many companies have to switch to models from non-European providers.
The SOOFI “Sovereign Open Source Foundational Models for European Intelligence” project is developing open, sovereign AI foundational models for industrial applications. The focus is on large multilingual language models with 70 to 100 billion parameters, domain-specific reasoning models and scalable multi-agent systems. The aim is to provide resilient, controllable and efficiently integrable AI technologies in line with the AI Act and European values.
In parallel, SOOFI is working on a multi-agent framework that combines large language models with tools, data spaces, digital twins and cloud edge infrastructures.
The technologies developed will be tested in use cases in medicine, production and logistics processes and administration in order to assess their usability and scalability.
