Mistral Releases Weights for Trillion-Parameter AI Model Built for Cybersecurity

French AI company Mistral released the weights for its trillion-parameter open-source model, Mistral Large 4 (ML4), on October 29, 2026, allowing businesses and governments to download and deploy the system independently. The model, nicknamed “Le Chonk,” had been introduced earlier that month.

With the weights now publicly available on Hugging Face and via API, organizations can run ML4 on their own infrastructure. Mistral says the release is aimed squarely at cybersecurity, enabling enterprises and government agencies to detect, analyze, and respond to threats across code and IT systems without depending on a third-party AI provider.

Pierre Stock, Mistral’s Vice President of Science, told TechCrunch that open-weighting the model gives organizations a form of access that “can’t be cut off by definition.” He added that companies can deploy ML4 on-premises within their own ecosystems and use it to “identify, coordinate, and respond in real time to major security incidents.”

Mistral claims ML4 performs comparably to GLM-5.3, a Chinese open-weight model previously considered among the best in the world for cyber-related tasks, while using 47 percent fewer tokens per task. The company also says ML4 ranks first among open models on Artificial Analysis’ Coding Agent Index for agentic coding tasks.

CEO Arthur Mensch framed the release around Mistral’s broader position that AI should “be a tool for empowerment, not dominance,” arguing the model achieves competitive performance at a fraction of the cost of closed-model alternatives.

Stock acknowledged concerns about rogue AI agents but cautioned against overreaction. “We don’t want to fearmonger people. We think it’s a problem that is solvable,” he said. He also emphasized the importance of making advanced defensive AI capabilities widely available, noting that malicious actors may already be leveraging similar technology to launch attacks.