Europe has taken a major step towards digital sovereignty by launching OpenEuroLLM, an initiative to create open source large language models (LLMs) for all European languages. The project, backed by the EU’s Digital Europe Programme, aims to ensure Europe’s AI development remains independent from foreign tech giants. OpenEuroLLM is a collaboration between 20 organizations, including academic institutions and tech firms, with a budget of €37.4 million dedicated to model development.
Led by Jan Hajič from Charles University in Prague and Peter Sarlin, CEO of Silo AI, the project will leverage high-performance computing resources from EuroHPC supercomputer centers in Spain, Italy, Finland, and the Netherlands. While the initiative aligns with Europe’s push for technological autonomy, some question whether a large consortium can compete with private AI firms like OpenAI or France’s Mistral AI.
Hajič emphasized that OpenEuroLLM is not starting from scratch. It builds on the High Performance Language Technologies (HPLT) project, which has already developed datasets and AI models using vast computational resources. The first LLM versions are expected by mid-2026, with final iterations by 2028.
The project aims to create multilingual foundation models that support the EU’s linguistic and cultural diversity while maintaining transparency. However, challenges remain, such as ensuring equal AI proficiency across all languages and addressing copyright issues surrounding training data. While OpenEuroLLM aspires to be fully open source, some data may be restricted under EU regulations.
This initiative follows a similar project, EuroLLM, raising concerns about duplication of efforts. However, OpenEuroLLM’s leaders remain optimistic, highlighting their strong funding and access to Europe’s supercomputing power. The ultimate goal is to provide European businesses and researchers with AI models built entirely within the continent, reducing reliance on foreign AI systems.
