Mistral unveils new AI model it says rivals best open systems from China (www.cnbc.com)

Mistral has launched its new flagship AI model, Mistral Large 4 (ML4), positioning it as the most capable open model outside of China and a significant competitor in the global AI landscape. With 1 trillion parameters, the model excels in various domains, including cybersecurity, coding, manufacturing, and finance. This move comes as Western developers seek to close the capability gap with their Chinese counterparts, who have historically dominated the open-weight and open-source AI sectors. Mistral's recent €3 billion funding round underscores its status as one of Europe's leading AI startups. ML4 was trained on 4,000 Nvidia Grace Blackwell GPUs over two months within Mistral's European data centers, promising top-tier performance on benchmark tests when its full parameters are released. Though ML4 still has some limitations, particularly in coding, the model is expected to improve as Mistral scales its training capacity post-funding. By promoting open AI models that can be modified and self-hosted, Mistral aims to empower enterprises and governments, especially against cyber threats that exploit vulnerabilities in closed systems. This further emphasizes the growing importance of open AI models in global cybersecurity scenarios.