Italy and Germany collaborate on a 400-billion-parameter open-source AI model to reduce reliance on the U.S. and China.

ME AI message, according to monitoring by Beating, the Italian AI startup Domyn has announced plans to release a fully open-source, reproducible large model with over 400 billion parameters within one year. The EUROPA consortium, formed by Domyn in collaboration with Germany’s Fraunhofer-Gesellschaft, has been selected for the European Commission’s Frontier AI Grand Challenge project and is receiving computing power support from the European public supercomputing network EuroHPC. Reducing dependence on overseas hosted systems is a core motivation for Europe’s pursuit of local alternatives. The open-source AI ecosystem is primarily led by Chinese companies, while leading U.S. models are mostly closed-source and require remote access. Due to concerns over U.S. export controls on large models like Anthropic, as well as restrictions imposed by Italy and the Czech Republic on DeepSeek’s remote services, European enterprises and governments are accelerating their shift toward localized systems that can be freely deployed on local infrastructure. Training data will be collected from institutional partners, with the first government data-sharing agreements expected to be finalized within weeks following meetings with several European heads of state. Regarding computing requirements, Domyn’s CEO Uljan Sharka noted that training a frontier large model from scratch consumes significantly less computing power than providing remote inference services to hundreds of millions of users, and the European public computing network is already sufficient to meet R&D needs. Funding for the project has been supported by G42 of Abu Dhabi, Eurizon Capital, Rabobank, and BNY Mellon. (Source: BlockBeats)