Mistral AI has unveiled its latest advancement in artificial intelligence, the Mistral Large 4 model, also known as “Le Chonk.” This 1-trillion-parameter multimodal model is positioned to redefine the landscape of AI technology with its enhanced capabilities in coding, cybersecurity, and reasoning efficiency. Now available with open weights, this model not only leads the pack in coding benchmarks but also offers organizations the flexibility to customize and deploy AI in a way that aligns with their strategic objectives.
Le Chonk is fully open-weight and trained on European infrastructure, underscoring its commitment to sovereignty in AI. This multimodal model has already shown remarkable performance, surpassing existing models in agentic coding and defensive cybersecurity. It provides three distinct reasoning modes—low, medium, and high—to cater to different task complexities, striking a balance between performance and cost.
Mistral Large 4 is redefining the standards of cybersecurity, achieving a score of 51% on the Artificial Analysis Cyber Index, highlighting its capability to identify, reproduce, and patch vulnerabilities without hindering current functionalities. Its refusal rate for harmful cyber requests stands at an impressive 96.6%, displaying its robust security guardrails.
Feedback from cybersecurity leaders such as Booz Allen, Cisco, and Cloudflare highlights Mistral Large 4’s advanced capabilities and customization potential. As organizations navigate the AI-driven security landscape, open-weight models like Mistral Large 4 offer unprecedented transparency and control. This model empowers enterprises to not only enhance their cybersecurity measures but also build resilient AI systems that operate independently of third-party services.
On the coding front, Le Chonk commands leading scores in benchmarks like FrontierCode 2.0 and Terminal-Bench 4.0. Offering superior token efficiency, it reduces the computational load and cost associated with developing complex coding agents. This makes Mistral Large 4 an economical choice for developers aiming to leverage sophisticated AI models in their work.
Beyond its prowess in coding and cybersecurity, Mistral Large 4 extends its utility to finance and other enterprise sectors. It boasts superior abilities in executing real-world workflows, such as spreadsheet management in finance scenarios. Its Finch score of 73.3% attests to its strong performance compared to other open models, making it invaluable for organizations focused on finance-related tasks.
The Mistral Large 4 model is available for deployment in customizable formats, further enabling organizations to align AI technology with their own infrastructure policies and data sovereignty requirements. Available under a modified MIT license, the model’s open-weight nature allows businesses to modify, fine-tune, and even commercialize the technology, bringing AI sovereignty to the forefront of enterprise innovation.
As it moves into broader availability, Mistral Large 4 serves as the cornerstone for further specialized and optimized models from Mistral AI. Enterprises keen to harness the potential of open-weight AI have a promising path forward with this robust model, leading to more adaptive, secure, and efficient AI-driven solutions across various sectors.
For those interested in exploring Mistral Large 4’s comprehensive capabilities, from cybersecurity to enterprise applications, additional information and access can be found by visiting the original article.
You can read the original article here: [Mistral AI’s official blog](https://mistral.ai/news/mistral-large-4-oss/)
