Top AI Product

Europe’s answer to the American and Chinese frontier labs now has a name: EUROPA. On June 19 the European Commission named the EUROPA consortium, led by Italian firm Domyn, winner of its Frontier AI Grand Challenge — and with it, the mandate to build an open-source frontier model in all 24 official EU languages.

## What’s being built

The target is a model exceeding 400 billion parameters, trained entirely on European public supercomputers. The consortium gets up to 2.5% of all EuroHPC capacity for a year, one of the largest AI compute allocations the EU has ever granted. The model will be permissively licensed and governed under the AI Act from the start.

## Why it matters

This is a sovereignty play, not just another big model. A permissively licensed frontier model trained on EU compute gives regulated industries — banks, healthcare, government — a deployment option that sidesteps vendor lock-in and the data-residency worries that come with closed American APIs. The 24-language scope is the other tell: most frontier models are English-first, and Europe is betting that native coverage of its own languages is worth building from scratch rather than fine-tuning in later.