Stay informed with free updates
Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.
French start-up Mistral has released a new series of powerful “open” artificial intelligence models, as it pushes to gain a crucial edge in the race to develop the cutting-edge technology.
The Paris-based group on Tuesday claimed its latest large language system, called Mistral Large 3, was one of the world’s best multimodal and multilingual open-weight models, offering the ability to be used in most European languages as well as some Asian ones.
The new release comes amid growing concerns that Europe is falling behind the US and China in AI development. Research last month showed that Chinese developers such as DeepSeek and Alibaba have overtaken their US rivals in the global market for open AI models for the first time this year.
Guillaume Lample, co-founder and chief scientist at Mistral, said making these models open was “aligned with our mission of putting AI in the hands of everyone”. He added: “I think it’s important that we have a company in Europe that is able to create this frontier model.”
Mistral has long pitched its model and chatbot as a more efficient and open-source alternative to US rivals such as OpenAI’s ChatGPT and Google’s Gemini. The two-year-old company is seen as Europe’s main hope for a homegrown player that can compete in the fast-developing strategic technology.
In contrast, the biggest US tech groups have opted to focus on “closed” models, preferring to maintain full control of their most advanced technology — profiting from them through customer subscriptions or enterprise deals.
“There is pretty much no alternative for Europe now to compete besides open sourcing,” said Lucie-Aimée Kaffee, EU policy lead and researcher at open-source start-up Hugging Face.
The release comes as the European Commission has called for greater AI sovereignty in a bid to decrease the bloc’s dependence on foreign technology providers. In October, it unveiled a new strategy aimed at promoting European AI tools to provide security and resilience while boosting the continent’s industrial competitiveness.
Mistral has also been lobbying European leaders to increase public funding, computing power and access to data for European open-source developers.
The French AI start-up is also releasing a series of smaller and faster open models, which can be used offline and offer users control over data as they do not have to go through Mistral’s online platform. This could enable them to be used for sensitive use cases, such as drones, robotics, healthcare and defence.
Hugging Face’s Kaffee said that Mistral’s decision to release powerful models on an open-weight basis placed Europe on the map for development of the technology.
“If you are an AI developer, you don’t have to go abroad. You can stay in Europe to create a company and to create competitive AI innovation,” Kaffee said. “Mistral has a very good position in the European system and the European space to really make the sovereignty angle of Europe a lot stronger.”
But others have criticised the company’s decision to opt for open-weight models, which are free to access but provide less comprehensive information than “open source” models that supply the code and training data required to train a model from scratch.
“Data at scale is really the missing key right now in the European AI innovation ecosystem and Mistral does not contribute to that at all,” said Andreas Liesenfeld, assistant professor at Radboud University in Nijmegen and the co-founder of the European Open Source AI Index.
