France2030: Opsci and the OpenLLM-France consortium winners of the ‘Digital Commons for Generative AI’ tender

The OpenLLM consortium is developing a sovereign and open-source multimodal generative AI. We believe in the importance of transparency and sharing in the development of AI technologies, and here is our manifesto.

The ambition is to create trusted, controlled, transparent, and genuinely open-source generative AI digital commons, meaning with training data available under open licenses. The goal is to provide a foundational model pre-trained on academic and high-quality French data, offering an API and a chat platform for education that leverages Open Educational Resources (OER). The use case aims to provide the essential digital commons to support AI implementation initiatives within the French Ministry of Education while enabling the edtech sector to access their roadmap for adopting these technologies, which allow for more personalized, interactive, and engaging learning paths to better train future generations.

In the face of the concentration of big tech players in AI and the current overcapitalization strategies being deployed, OpenLLM seeks to be a collaborative, open, and industrial alternative, creating digital biodiversity and commons with a focus on the public interest and accessibility for the greatest number.

The strategic decision by the French government, which announced the winners of the France 2030 call for projects during the Gathering of the Best French AI Talents organized by the Presidency of the Republic, demonstrates its support for our collaborative and open approach, as well as recognition of the work already done by the consortium on the Claire and Lucie models.

Alongside Opsci, the consortium is led by Linagora and includes: GENCI and IDRIS of CNRS, LORIA, CEA-List, Université Paris 1 Panthéon-Sorbonne, LIX from École Polytechnique, Association Class’Code, and the company Talkr.ai.