After the public preview at the beginning of October, the French company Mistral AI has made Mistral Large 4 available to everyone, commercial name «Le Chonk», the latest version of its artificial intelligence model. The model is designed to compete with AI solutions developed by tech giants from the United States and China. The initiative represents one of Europe’s main attempts to assert its autonomy in the digital sector, offering an alternative to dependence on foreign infrastructure to give continental organizations control over sensitive data.
An “open-weight” model to install on your servers
The new system – as explained on the company website – adopts an “open-weight” approach, i.e. it makes the fundamental parameters available to developers to allow the installation and management of software on proprietary servers, without the obligation to send information outside European borders. Trained on a network of thousands of advanced processors, according to Mistral AI the model can understand and generate texts in over one hundred and sixty languages and is structured to process complex calculation and automation tasks. One area of interest is cybersecurity, in which OpenAI, Anthropic and Microsoft have invested in the recent past with specific versions of their models designed for cybersecurity tasks.
Code testing and abuse protections
In comparative tests on code generation, the French solution surpasses the main Chinese models, such as DeepSeek, while in the tests of identifying flaws and writing corrections, it ranks close to the global giants. To prevent improper use or cyber attacks, Mistral announced that it has implemented automatic protection systems, capable of intercepting and blocking potentially harmful requests.