Mistral Unveils Large 4, a One Trillion Parameter AI Model Focused on Cybersecurity

Mistral has launched its largest AI model to date, Large 4, featuring one trillion parameters. According to The Decoder, this model is trained entirely on European infrastructure, marking a significant step for AI development outside the US and China.

The Decoder reports that while Large 4 shows notable progress on the independent Intelligence Index, it still trails behind leading models such as Claude, GPT-6, and some Chinese competitors. Mistral’s key differentiation lies in targeting cybersecurity tasks that US-based AI models tend to avoid.

This development may attract attention from Japanese markets, where cybersecurity remains a critical concern amid increasing digital threats and a growing interest in locally compliant AI solutions.