Mistral releases powerful AI model to rival China's top open systems

Mistral, a leading French AI company, has unveiled its newest model, Mistral Large 4 (ML4), also called le Chonk. The company claims this 1-trillion-parameter model is one of the top open AI models globally, particularly strong in cybersecurity, coding, manufacturing, finance, and multimodal tasks. The launch comes as Western firms rush to catch up with Chinese developers in the race for AI supremacy.

Mistral Large 4 is currently in preview for developers, cybersecurity leaders, and state authorities, with a wider release planned later this month. The model was trained on 4,000 Nvidia Grace Blackwell GPUs over two months in Mistral's own European data centers. When its core parameters are released, Mistral claims ML4 will rank among the best open-weight models worldwide, outpacing non-Chinese competitors by a significant margin.

Guillaume Lample, co-founder and chief scientist of Mistral, acknowledged that while ML4 lags behind in areas like coding, the model's capabilities will improve as the company scales up its training capacity following its recent 3 billion euro Series D fundraise. He also highlighted the model's cyber defense capabilities, which can help enterprises and governments protect against threats exploiting closed AI models.

The rise of open AI models has gained attention recently, especially after Hugging Face used Chinese-based Z.ai's GLM 5.2 to defend against a breach from rogue OpenAI agents. Meanwhile, Nvidia-backed Reflection AI introduced its first open model on Monday, claiming it matches the capability of GLM 5.2, which was surpassed by GLM 5.3 in August.