French AI developer Mistral AI has announced the open public release of its flagship model, Mistral Large 4 (also known by its working name Le Chonk). The 1 trillion parameter model is natively multimodal and designed to detect and remediate cybersecurity vulnerabilities, as well as solve complex autonomous coding problems.
French AI developer Mistral AI has announced the open public release of its flagship model, Mistral Large 4 (also known by its working name Le Chonk). The 1 trillion parameter model is natively multimodal and designed to detect and remediate cybersecurity vulnerabilities, as well as solve complex autonomous coding problems.
Unlike leading closed systems like Google Gemini or OpenAI, which are accessed exclusively through cloud APIs, Mistral Large 4 weights can be freely downloaded and deployed locally on your own infrastructure.
Recall that Mistral AI introduced Large 4 in early October. According to the independent rating of Artificial Analysis, it became the smartest AI model created outside the US and China.
The public version of the model has received improvements compared to the previous announcement three weeks ago, the company announced on its blog. The changes were made based on feedback from government agencies and corporate customers who tested the model in early access mode.
With the ability to deploy locally, companies and government agencies can use AI tools on their own servers without sharing sensitive data with third parties. Mistral says this is their most powerful open model for cybersecurity, and could compete in a segment currently dominated by Chinese developments.
“I feel uncomfortable living in a world where there are no open-source models. Once a model is out there, it can no longer be restricted. This protects users from the risk of being disconnected from the service, while also meaning that anyone can use it,” said Pierre Stock, Vice President of Science at Mistral AI.
