Mistral AI Releases Le Chonk Open-Weight Model for Cyber Defense
Mistral AI released Mistral Large 4, a trillion-parameter open-weight model designed to provide Western governments and enterprises with advanced agentic coding and cybersecurity capabilities.
French AI company Mistral AI released Mistral Large 4, a trillion-parameter open-weight model nicknamed Le Chonk. Launched on Thursday following a public preview in October, the final version incorporates reinforcement learning and fine-tuning based on feedback from developers and cybersecurity experts. The model is designed for multimodal processing of text and images, supporting over 160 languages and geospatial analysis of satellite imagery.
Mistral AI positions the model as the world's best open-weight offering for automated coding and defensive cybersecurity. It outperformed Opencode's GLM-5.3 on the Terminal-Bench 4.0 benchmark and scored above DeepSeek V4.1 Flash on the CyberGym benchmark. The company claims the model is more efficient than Chinese rivals, using 47% fewer tokens per task on average. Despite these gains, proprietary models like Anthropic's Mythos 5.1 continue to lead in overall coding performance.
Company leadership describes the release as a strategic move for European data sovereignty. By providing open weights, Mistral AI aims to ensure business continuity for enterprises and governments, removing reliance on proprietary U.S. models that can be remotely disabled or restricted by export controls. While the company emphasizes these security benefits, some security experts warn that the model's dual-use nature could allow threat actors to customize it for malicious purposes.
