Mistral Large 4 is now available as an open-weight AI model, allowing businesses and governments to run it on their own infrastructure. The French company released the weights Thursday, emphasizing cybersecurity, software development and independent AI deployment.
Nicknamed Le Chonk, the trillion-parameter model debuted in a public preview in early October. Its downloadable weights are now available on Hugging Face, and it’s also accessible through Mistral’s API.
Mistral says organizations can use the system to detect vulnerabilities, analyze security incidents and respond to threats without relying entirely on external AI providers.
Mistral Large 4 Targets Cybersecurity and Coding
The company claims its latest model delivers competitive performance against leading open-weight systems, including China’s GLM-5.3.
According to Fortune’s reporting, Mistral says Large 4 uses 47% fewer tokens per task than GLM-5.3. The company also reports leading performance among open models on Artificial Analysis’ Coding Agent Index.
These efficiency improvements could reduce computing costs for organizations running complex AI workloads. However, benchmark results do not guarantee the same performance in real-world applications.
For cybersecurity teams, the model supports vulnerability analysis, malware investigation and incident response. Organizations can deploy it within their own environments while maintaining greater control over sensitive information.
Mistral Vice President of Science Pierre Stock said open weights let companies keep access to advanced AI capabilities. Unlike services controlled entirely by external providers, downloaded models can remain available during service restrictions or outages.
Open-Weight AI Raises Security Questions
The release strengthens Mistral’s position against closed-model developers such as OpenAI and Anthropic. CEO Arthur Mensch has argued that advanced AI should empower organizations rather than concentrate control among a few providers.
That position reflects a broader industry debate over open-weight AI, particularly its role in cybersecurity and national security.
Stock acknowledged that powerful AI systems can also create risks when malicious actors gain access to similar capabilities. However, he argued that defenders need advanced tools to identify vulnerabilities and respond to increasingly sophisticated attacks.
Mistral maintains that organizations should be able to operate cybersecurity models under their own policies. The release gives enterprises and government agencies another option to deploy advanced AI without relying exclusively on proprietary cloud services.
