The Jean ZAY supercomputer puts French Artificial Intelligence research centre stage!

Faced with growing demand from AI communities (+1,000 research projects by 2023) and with the rise of Generative AI, GENCI is announcing a €40m investment to extend the supercomputer's capabilities. 

Key figures for this expansion:

• 1456 NVIDIA Hopper GPUs in 14 BullSequana AI 1200H racks  ;
• A 4.3 PB storage infrastructure, offering 1.2 TB/s of bandwidth;
• A French design with Eviden for a perfect installation in just 4 months!

The development of this score has enabled us to serve 13 major projects since the summer of 2024, including LUCIE, the 1st truly Open Source generative AI model, with 100% transparent training data.

LUCIE is a multimodal foundation model with 7 billion parameters. Trained on Jean ZAY with 512 GPUs for 3 months, it aims to support the creation of educational content, initially in French, then in 4 other European languages.

With a high-quality database of 3 trillion tokens and a state-of-the-art tokenizer, this model will become a benchmark for producing quality content for learners across Europe.
 

" The formation of LUCIE marks a key step towards sovereign AI [...] With LUCIE, LINAGORA and the OpenLLM France / Europe consortium are committed to providing a truly open source core model, which is a major step forward in AI for education that can be extended to multiple languages and domains, including coding and mathematics [...].. "

Michel-Marie MAUDET, Managing Director and Co-Founder of LINAGORA

Its advances symbolise the synergy between computing power, AI and European ambition. A tool at the service of science, innovation and technological sovereignty.

+ Find out more about LUCIE :
Webinar Replay
Détails Details of training phases