400 Billion Parameter Model: Consortium "Europa" Wins AI Competition

The EU Commission selects a consortium led by the Italian company Domyn as the winner of the "Frontier AI Grand Challenge" to train a European AI model.

listen Print view
EU flag waving against a blue sky.

(Image: Andy.LIU / Shutterstock.com)

4 min. read

The EU Commission announced the winner of its “Frontier AI Grand Challenge” on Friday. The consortium “Europa”, led by the Italian company Domyn, prevailed in the competition. The alliance is to receive the necessary resources with the award to develop a state-of-the-art open-source AI model. The prestige project is intended to cover all 24 official EU languages and set a statement for the continent's technological ambitions.

With this decision, the Commission wants to strengthen European sovereignty in the field of cutting-edge technology. According to the Commission, Europe has the necessary talent and industrial capacity to play a leading role in the global AI race.

The initiative, launched in February of this year, called on the continent's leading AI innovators to design a system with more than 400 billion parameters. Until now, such a scale has been almost exclusively reserved for the most advanced and financially strongest models from the USA and China.

The technological basis of Domyn, which the AI forge is advancing in cooperation with Nvidia as part of the European Sovereign AI initiative, also operates in this league. Common language models are primarily optimized for academic targets but often neglect the specific needs of strictly regulated sectors such as financial services, defense, or advanced manufacturing.

These industries demand AI systems specialized for business-critical workflows, such as querying structured databases. The consortium leader is already addressing this with the 263 billion parameter model Domyn-Large, as well as its predecessors Italia-10B and Colosseum-355B, which are available as Nvidia NIM microservices and optimized for European core languages in accordance with the AI Act.

The follow-up project presented as part of the competition aims to create an even more powerful frontier model. These are extremely versatile systems that can be transferred to a wide variety of application areas with minimal adaptation effort. To efficiently handle the enormous computational load, the developers rely on modern, modular architectures such as the “Mixture-of-Experts” method. This is intended to set benchmarks in terms of performance and resource efficiency.

The final model will be made freely available to the public as open-source software. This is intended to ensure that companies, researchers, and public institutions in the EU can benefit equally from the progress.

Videos by heise

Financing and infrastructure are guaranteed through the Commission's cooperation with the supercomputing consortium EuroHPC. The winning team will have access to up to 2.5 percent of the total EuroHPC computing capacity on one or more European supercomputers optimized for AI applications for one year. This technological boost is part of a broader strategy within the “Action Plan for a European AI Continent” to strengthen the European startup ecosystem and establish the community as a leading AI location.

Commission Vice-President Henna Virkkunen emphasized the claim that the EU wants to take a leading role in advanced AI. Project Europa proves that one can keep up with the best in the world while remaining true to one's own values. The model will stimulate innovation in key areas such as manufacturing, healthcare, and autonomous systems and consolidate a trustworthy AI ecosystem “Made in Europe.”

(vbr)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.