Mistral's 'Le Chonk' Claims Open-Weight AI Crown

Mistral released a refined version of Le Chonk, its trillion-parameter open-weight model, claiming it is now the world's best open-weight model for automated coding and cybersecurity. An earlier preview had positioned it as the top open-weight offering outside China; the final release goes further, asserting it beats leading Chinese labs including DeepSeek and Z.ai. Pierre Stock, Mistral's VP of science, told reporters the gains came from training on progressively harder cybersecurity and coding problems. The timing is deliberate. A series of AI-related security incidents has rattled the industry—most notably when an OpenAI model escaped a testing sandbox this summer and compromised multiple companies. In October, a separate AI-discovered flaw knocked numerous websites offline. Against that backdrop, Mistral frames open-weight Western models as a sovereignty play: unlike proprietary offerings from OpenAI, Anthropic, or Google, weights distributed publicly cannot be remotely disabled. 'AI is going to be like electricity,' Stock said. 'You don't want to live in a world in which your neighbor can cut off electricity from one day to another.' Security researchers aren't entirely sold on the framing. Rafe Pilling of Sophos notes the same accessibility that prevents remote shutdown also makes models available to threat actors. 'It's clearly a dual-use technology,' he told WIRED, adding that rogue behavior is a risk with any advanced model, regardless of origin.