OpenEuroLLM secures strategic access to multiple supercomputers: A milestone for European language models

The EU has outlined a clear roadmap to secure Europe’s technical and economic independence in AI. A central component of this strategy is access to world-class computing power. OpenEuroLLM is now the first AI project to be granted strategic access to several of EuroHPC’s supercomputers simultaneously, including LUMI, Leonardo, Jupiter, and MareNostrum 5.

These resources will be used to train and develop a family of high-performance, open-source language models for all official EU languages. The decision secures the infrastructure to build next-generation, transparent and legally compliant language models that reflect Europe’s languages and cultural diversity.

Transparency as a competitive advantage

Developing the models as open source is a strategic necessity for digital sovereignty. It enables full transparency and control—factors that are crucial for Swedish companies and the public sector to implement AI that is both competitive and compliant with European regulations.

The project brings together 20 leading partners, including EuroHPC centers and research institutes. AI Sweden’s participation rests on a solid foundation of experience—from the development of the first large Nordic language model, GPT-SW3, to the ongoing work within TrustLLM, which focuses on reliable models for Germanic languages.

"No single actor can match the resources required to train models of this size. By having AI Sweden as part of OpenEuroLLM, we can help connect the Swedish ecosystem to absolute world-class resources," says Magnus Sahlgren.

The work to utilize these new resources begins immediately as part of the project’s three-year plan to strengthen Europe’s AI capacity.