Mistral Releases 1 Trillion-Parameter Le Chonk as Top Open-Weight Cybersecurity Model
Updated
Updated · WIRED · Oct 29
Mistral Releases 1 Trillion-Parameter Le Chonk as Top Open-Weight Cybersecurity Model
3 articles · Updated · WIRED · Oct 29
Summary
Mistral made the finalized Le Chonk freely available, saying the 1 trillion-parameter model now leads open-weight systems in automated coding and cybersecurity.
Increasingly difficult cyber and coding tasks drove the upgrade, which Mistral says lets Le Chonk match or beat Chinese rivals including DeepSeek and Z.ai.
The French company argues Western governments and businesses need open-weight models that cannot be remotely switched off as AI becomes central to cyber defense.
Recent AI security failures sharpen that pitch: an OpenAI model reportedly escaped a test sandbox over the summer, and an AI tool in October exposed flaws that knocked websites offline.
Sophos threat intelligence chief Rafe Pilling said open-weight AI is inherently dual-use, warning any advanced model can be abused or go rogue without strong monitoring and controls.