France’s Mistral unveils latest model in sovereign AI push

Mistral said Chonk’s scores on performance benchmarks so far make it one of the best open-weight models available and “by far the best in Europe and the U.S.” A phone folder with AI chatbots. (Solen Feyissa/Unsplash via Courthouse News) (AFP) — French AI startup Mistral announced its biggest open-weight model Tuesday, dubbed “Chonk,” claiming enhanced performance and security capabilities for clients seeking to keep their data private instead of relying on closed models from companies such as Anthropic and OpenAI. The new model lets users download and customize the parameters, or “weights,” that determine how it processes information and run it on their own servers. It is Mistral’s first major product launch since it raised 3 billion euros ($3.4 billion) last month in Europe’s biggest technology fundraising ever. The company is betting that businesses and governments will increasingly want total control over their AI deployments, as opposed to using proprietary models that require handing over data to third parties and can be much more expensive. “Enterprises across financial services, manufacturing and the public sector can govern the intelligence, data, compute and operations underpinning their critical infrastructure … without exposing their most valuable knowledge to anyone outside their own walls,” the company said in a statement. “That one took some groundwork,” Mistral CEO Arthur Mensch posted on X. “Trained and served on our own compute, and RL shows no sign of saturation,” he wrote, referring to reinforcement learning, a method of further developing AI models, and whether additional training produces diminishing performance returns. Open-weight models are also touted as more secure, a growing concern after several incidents of AI agents going rogue by hacking databases or otherwise doing things they were not supposed to do. “The cyber defence capabilities will enable enterprises and governments to defend themselves against threat actors that are jailbreaking closed models to perform cyberattacks,” Mistral chief scientist Guillaume Lample said in the statement. Data ‘under their control’ Mistral, considered Europe’s leading AI developer, was valued at more than 21 billion euros in last month’s funding round — far below the valuations for U.S. giants such as OpenAI or the AI operations of Google, Meta and SpaceX. Anthropic is reportedly preparing an initial public offering in the coming weeks that could value it at $2 trillion. But the companies’ closed models, which currently dominate the market, are facing growing competition from open-weight models, largely from Chinese developers, including DeepSeek and Moonshot. These models can be fine-tuned by clients for specific needs, run on in-house infrastructure and used at a fraction of the cost. Mistral’s new model also comes as a U.S. open-weight rival, Reflection AI, announced its widely anticipated Beam model Monday, also promising cheaper operating costs. Mistral said Chonk’s scores on performance benchmarks so far make it one of the best open-weight models available and “by far the best in Europe and the U.S.” It said the model, which is still being refined, was built using 4,000 Nvidia chips over two months at its own data centers in Europe. Customers can either download the model to their own servers or access it through Mistral’s infrastructure, including in “our European sovereign region where data stays under EU jurisdiction to ensure data remains under their control.” Mistral said the model was trained on more than 160 languages and can answer prompts and questions in each one natively. Our weekly newsletter Closing Arguments offers the latest about ongoing trials, major litigation and rulings in courthouses around the U.S. and the world, while the monthly Under the Lights dishes the legal dirt from Hollywood, sports, Big Tech and the arts.