French startup Mistral has unveiled Europe's first AI model designed specifically for reasoning, aiming to challenge leading American and Chinese players in the artificial intelligence sector. This new model generates responses using logical and structured thought processes. Mistral, backed by French President Emmanuel Macron, is setting itself apart by emphasizing its European identity and by releasing some of its models as open source—unlike major players such as OpenAI and Google, which primarily keep their models proprietary.

Although Mistral is seen as Europe’s strongest hope for a homegrown AI leader, it still trails its global counterparts in terms of revenue and market share. The company’s new reasoning models apply a "chain-of-thought" method, solving complex problems step-by-step using intermediate reasoning. This technique could be key in enhancing AI's performance, especially as the strategy of simply scaling up model size with more data and computing power shows diminishing returns.

Valued at $6.2 billion by investors, Mistral could benefit from the industry's growing interest in innovation over pure scaling. Earlier this year, China’s DeepSeek made significant progress by releasing low-cost, open-source reasoning models. OpenAI led the way with reasoning models in 2023, followed by Google, while Meta, although active in open-source AI, has not yet launched a dedicated reasoning model. Mistral’s latest release includes the open-source Magistral Small model, along with a more powerful Magistral Medium version intended for enterprise users.

“Human thinking isn’t always linear—it blends logic, intuition, uncertainty, and discovery,” the company said. “Reasoning language models allow us to delegate and enhance complex thought through AI.” While U.S. firms typically keep their most advanced models closed, some like Meta have opted for open-source releases. Chinese companies including DeepSeek and Alibaba are increasingly using open-source strategies to showcase their tech capabilities.