Mistral releases weights for its trillion-parameter AI model, Mistral Large 4
French AI company Mistral has released the weights for its trillion-parameter Mistral Large 4 model. This model, which organizations can run on their own infrastructure, aims to provide independence, particularly in cybersecurity defense.
French AI startup Mistral has released the weights for its trillion-parameter AI model, Mistral Large 4 (ML4), codenamed "Le Chonk," to the public. This move allows businesses and government agencies to host the system on their own servers, enabling them to conduct cybersecurity defense without relying on external providers.
Mistral CEO Arthur Mensch stated that AI should be an empowering tool, noting that the new model aims to deliver performance comparable to closed-source labs while using significantly less capital and compute power. The company claims that ML4 is comparable to the GLM-5.3 model but uses 47 percent fewer tokens per task.
The model's weights are available via the Hugging Face platform and API. Pierre Stock, Mistral's VP of Science, emphasized that this release strategy provides organizations with full control over AI-powered cyber defense systems and eliminates the risk of access being cut off. The system can be used to identify, coordinate, and respond to security incidents in real time.
Sources
- techcrunch.com · Link