MISTRAL UNVEILS LARGE 4 AI MODEL TO RIVAL CHINESE SYSTEMS

Mistral has unveiled a new flagship artificial intelligence model, Mistral Large 4, which the French company says ranks among the best open models available outside China. The model, also referred to internally as "le Chonk", contains 1 trillion parameters and is described by Mistral as particularly effective at cyber, coding, manufacturing, finance and multimodal tasks. The company says it is being launched first in preview to developers, cybersecurity specialists and state authorities, ahead of a wider release later this month. Mistral says the model was trained over two months using 4,000 Nvidia Grace Blackwell GPUs housed in its own European data centres.

Mistral, based in France, has been described as Europe's principal competitor to American developers OpenAI and Anthropic. The company raised 3 billion euros in a Series D funding round in September, valuing it at 21 billion euros, according to the report. Mistral says that once its core parameters are released, Large 4 will rank among the top open-weight models globally on aggregate benchmark performance, and that it is the strongest open-weight model developed outside China by what the company called a "substantial margin". Mistral acknowledges the model still lags behind the leading systems in areas such as coding.

The release comes as access to open AI models, which can be modified and self-hosted unlike closed systems, has become a point of competition between the United States and China, with Chinese-developed open models increasingly adopted internationally. A separate report said Nvidia-backed Reflection AI has announced its own open model, called Beam, which it describes as comparable in capability to Chinese models. Mistral has not said when it will release Large 4's core parameters to the public beyond confirming a wider rollout later this month.