Mistral releases weights for its trillion-parameter Mistral Large 4 model, pitching it for cyber defense
TechCrunch reports the French company says businesses and governments can now download and run the model themselves to detect and respond to threats.
French AI company Mistral has released the weights for Mistral Large 4 (ML4), its trillion-parameter open source model, according to TechCrunch. With the weights released, businesses and governments can download the system and run it themselves. TechCrunch reported that the model, nicknamed "Le Chonk," was first introduced earlier this month.
According to the report, the weights are available on Hugging Face and through Mistral's API.
Mistral is aiming the model at cybersecurity. "Enterprise and public sector institutions can now download, deploy, and own frontier AI to detect, analyze, and respond to threats across code and IT infrastructure," the company said Thursday.
Mistral says ML4 performs about as well as GLM-5.3, a Chinese open-weight model that TechCrunch said was previously considered one of the best open-weight models, especially for cyber work. Mistral says its model uses 47 percent fewer tokens per task. The company also claims ML4 ranks above all other open models for agentic coding tasks on Artificial Analysis' Coding Agent Index, an AI benchmarking site.
TechCrunch places the release within growing competition between open-weight developers such as Mistral and companies with closed models, such as OpenAI and Anthropic. Open-weight developers argue that businesses and governments should not have to depend on outside AI companies for advanced capabilities. TechCrunch said that argument may carry more weight in cybersecurity, where organizations want more control over their data and systems.
Mistral CEO Arthur Mensch has said AI should "be a tool for empowerment, not dominance," TechCrunch reported.
Pierre Stock, Mistral's vice president of science, told TechCrunch that open weights give organizations access to advanced AI for defense, access that "can't be cut off by definition because the model is open-weighted." He said companies could deploy the model on their own premises, and that it could be used to "identify, coordinate, and respond in real time to major security incidents."
Stock said it was important to give this capability to everyone as soon as possible, because malicious actors might use their own technology to attack. He also acknowledged the risks raised by recent incidents involving rogue AI agents, but cautioned against overreacting.
“We don't want to fearmonger people. We think it's a problem that is solvable. It's a hard problem to have good security.”Pierre Stock, Vice President of Science, Mistral
Several questions remain open. The performance and efficiency figures are Mistral's own claims, and the TechCrunch report does not describe any independent testing of them. The report does not give the license terms, the hardware needed to run the model, or which organizations, if any, plan to deploy it. It also does not describe the rogue AI agent incidents Stock mentioned.
