French AI startup Mistral AI has launched Mistral Large 4 (ML4), a new large multimodal model with 1 trillion parameters designed to compete on a global scale against leading American and Chinese AI systems. The company aims to carve out a “third way” in AI development, as described by French President Emmanuel Macron, positioning ML4 as a more transparent and accessible alternative to both closed models, which can be restricted or unplugged, and open models often dominated by China. Currently, ML4 is accessible only through a moderated public interface, but Mistral plans to release the model weights within a few weeks after completing safety evaluations.

Mistral emphasizes the control and security of its open-weight model, pledging collaboration with trusted partners and governments to prevent misuse while enabling transparency and auditability. According to Pierre Stock, Mistral’s VP of Science, this approach addresses recent heating security concerns among enterprise and institutional customers who require dependable, auditable AI tools. The company also points out that ML4’s training was accomplished using 4,000 Nvidia GPUs, which is substantially less compute power than competitors from China or closed-source players, highlighting an efficient use of resources.

While benchmark results are forthcoming, Mistral expects ML4 to lead among open models worldwide and hopes its specialized multimodal capabilities will surpass some closed models in domains critical to its target users. High-value use cases include cybersecurity, finance, and chip design. The latter is particularly noteworthy as Mistral counts industry giants like Dutch chipmaker ASML and tech conglomerate Samsung among its key investors, with Samsung leading a recent €21 billion valuation funding round.

Mistral is keen to differentiate itself beyond merely being an inference provider, continuing to position as a frontier AI research lab. The release of ML4 marks a significant milestone as European players continue to maintain a presence in the competitive AI landscape, offering alternatives to both closed American and open Chinese ecosystems. This strategic positioning highlights Mistral’s ambition to be a major force in shaping the future of sovereign and open AI technologies.