Mistral’s new Le Chonk model brings AI cybersecurity to your business – and you control it

Mistral has shipped its latest model, Mistral Large 4 (ML4), nicknamed "Le Chonk" for its trillion-parameter size, and positioned it as an open solution for cyber defense requirements. The model is available now in public preview, and Mistral said it will release the model weights on Oct. 27 — a gap the company said gives it a month to work with developers, cybersecurity leaders and state authorities to further assess ML4's capabilities and behavior in real-world environments.

Mistral said ML4 is built for security that stays under user control, unlike proprietary models, which it said frontier labs can technically rescind access to at any time. "The cyber defense capabilities will enable enterprises and governments to defend themselves against threat actors that are jailbreaking closed models to perform cyberattacks," Mistral co-founder Guillaume Lample said. The company said ML4 prioritizes cyber defense capabilities, customizable control and data sovereignty.

In a briefing, Lample and Mistral VP of Science Pierre Stock emphasized that security is a key requirement for the company's enterprise clients. Following the Hugging Face breach, Mistral was among the companies that signed Nvidia's Open Secure AI Alliance, a cross-industry partnership arguing open models are crucial to democratizing defenses against AI security incidents.

A Mistral spokesperson told ZDNET that "in absolute terms on cyber capabilities, ML4 outperforms the best models from Kimi, Deepseek and Meta." Mistral said initial testing partners will get access to a less guardrailed version of ML4 with expanded cybersecurity capabilities, similar to Anthropic's Project Glasswing and OpenAI's rollout of Astra.

Mistral said Le Chonk also excels in finance and multimodal use cases. It said it is still waiting on final benchmarks; early third-party analysis shows ML4 competing on par with GPT-6 Astra in certain computer vision tasks, meeting or slightly outperforming DeepSeek models on financial work tasks, and reaching a new high of 15% for open-weight models on Harvey's Legal Agent benchmark. Mistral said ML4 was trained from scratch on 4,000 Nvidia Grace Blackwell GPUs over two months in its own European data centers, and that the preview will run on those same GPUs. Nvidia CEO Jensen Huang said on X that OpenAI trained GPT-6 Astra on roughly 100,000 GPUs, according to the company. Stock said Mistral is "fully separate from other models, and we don't take inspiration from them," addressing accusations that Chinese labs have distilled proprietary models. Mistral said customers will soon have flexible deployment options, including self-deployment, API access in a region of their choice, and a European sovereign region where data stays under EU jurisdiction.