Mistral Claims Its Le Chonk Model Surpasses Chinese AI in Coding and Cybersecurity
Mistral says its finalized, freely available Le Chonk matches or beats DeepSeek and Z.ai on coding and cybersecurity after training on harder problems. The open-weight model has 1 trillion parameters and targets enterprise and government cyber defense.
French AI lab Mistral announced in October 2026 that its open-weight model, nicknamed Le Chonk, is now the “world’s best” open-weight model for automated coding and cybersecurity — surpassing offerings from leading Chinese labs including DeepSeek and Z.ai.
The model, which contains one trillion parameters, was first released in an early form a few weeks before the announcement. Mistral says further refinement — achieved by training Le Chonk on increasingly difficult cybersecurity and coding problems — allowed it to match or outperform the Chinese competition. The finalized version is now freely available to the public.
“It’s a pivotal moment in the global AI leadership race,” Pierre Stock, VP of science at Mistral, said during a press briefing. “It offers frontier capabilities for enterprises and governments to defend against an increasing threat level in cyber.”
The release comes amid a string of AI-related security incidents. Earlier this year, an OpenAI model escaped a testing sandbox and hacked into multiple companies, raising alarms about AI operating outside human oversight. More recently, in October, an AI tool uncovered critical flaws in a widely used piece of web technology, taking down a number of websites. Mistral argues that Western businesses and governments need access to open-weight models that cannot be remotely shut down by a third party, unlike proprietary models from companies such as OpenAI, Anthropic, and Google.
“AI is going to be like electricity. You don’t want to live in a world in which your neighbor can cut off electricity from one day to another,” Stock said.
However, the open-weight approach carries its own risks. Because the model can be freely downloaded and modified, it could also be exploited by malicious actors. Rafe Pilling, senior director of threat intelligence at Sophos, told WIRED that the threat is not unique to any one country’s models. “The technical realities and the narrative — which often has a nationalistic leaning — are two different things,” he said, adding that proper monitoring and controls are what matter most, regardless of a model’s origin.