OpenEuroLLM – building the next-generation open-source language models

The OpenEuroLLM project, an unprecedented collaboration, aims to propel European AI capabilities to new heights.

Share this news

In a groundbreaking move for the European artificial intelligence scene, top AI companies and research institutions have joined forces to develop next-generation open-source language models. The OpenEuroLLM project, an unprecedented collaboration, aims to propel European AI capabilities to new heights.

Led by a consortium of 20 renowned research institutions, companies and EuroHPC centers, OpenEuroLLM is set to build a family of performant, multilingual, large language foundation models. Coordinated by Jan Hajič (Charles University, Czechia) and co-led by Peter Sarlin (AMD Silo AI, Finland), the project will deliver cutting-edge AI solutions tailored for commercial, industrial and public sector applications. 

“In this project, co-led by Peter Sarlin (AMD Silo AI, Finland), we will build a fully open set of large language models that will cover all EU languages (and beyond), with as equal, as high quality as possible. These models will come in several sizes for easy deployment, and, of course, be compliant with the European regulations. We will show that this is possible with a transparent, collaborative effort, with the relatively moderate computing power available to us in Europe. Evaluation benchmarks will respect all the languages and the associated cultural aspects. All these features together will bring new-generation models, processes and datasets for European companies and further research efforts.” (Jan Hajič) 

Jan Hajič is part of the Data and AI innovation community of BDVA and was the main host and major contributor to the success of EBDVF 2022, our event in Prague. 

By prioritising transparency and compliance, these open-source models will democratise access to high-quality AI, empowering European businesses to compete globally and enhancing public services with advanced technology. The project offers a good example of the type of technology infrastructure needed to lower thresholds for European AI product development and refinement. OpenEuroLLM aligns with Europe’s strategic vision for digital sovereignty and competitiveness, reinforcing the importance of openness, collaboration and community-driven innovation, values that are the pillars of the BDVA community. 

“The project has a strong community-building component. The project will establish and Open Strategic Partnership Board with Open Science and Open Source advocates, experts in LLM building, data acquisition and evaluation. Everybody is welcome to suggest additional textual datasets to use, contribute to the community-based effort to help with evaluation datasets for language- and culturally-aware benchmarks and with their expertise.” (Jan Hajič) 

Developed within Europe’s robust regulatory framework, the models will uphold European values while achieving technological excellence. By collaborating with open-source communities such as LAION, open-sci and OpenML, alogside additional experts in the field assembled in the project’s Open Strategic Partnership Board, the project ensures that its AI models remain fully accessible and adaptable to diverse industry and public sector needs. With a strong emphasis on linguistic and cultural diversity, these performant multilingual models pave the way for impactful AI solutions, allowing European companies the possibility to develop high-quality products. 

Recognised with the prestigious STEP (Strategic Technologies for Europe Platform) seal, the project builds on the expertise and achievements of past European AI initiatives, including large repositories of high-quality data and pilot LLMs developed previously. Supported by the European Commission under the Digital Europe Programme, the consortium was officially launched on February 1st, 2025 – ushering in a new era of AI excellence for Europe and beyond. 

We are very proud to see that part of our community is at the core of this project, as most of the consortium consists of BDVA members. Congratulations to all and we look forward to your contribution to the next-generation of European open-source language models!

We are thrilled to see the incredible participation of BDVA members in this shared endeavour! Together, we look forward to shaping the next generation of European open-source language models, as we will join our efforts in this exciting journey! 

 

Full list of partners 

Universities and Research Organisations: 

Companies: 

EuroHPC centres: 

  • SURF, the Netherlands