French AI company Mistral has released the multi-trillion parameter open source AI model Mistral Large 4 (ML4) Weights, allowing businesses and governments to download and deploy the system themselves. The model, nicknamed “Le Chonke,” was first introduced earlier this month.
The release comes amid increasing competition between openweight AI developers like Mistral and closed-model providers like OpenAI and Anthropic, with the former arguing that businesses and governments should not rely on outside AI companies to access advanced capabilities. This discussion may be more relevant in cybersecurity, where organizations want more control over their data and systems.
Or, as Mistral CEO Arthur Mensch explained, AI should be “a tool for empowerment, not domination.” With Chonk, Mistral further strengthens its position, claiming it can achieve “state-of-the-art performance” with “a fraction of the computational power and capital invested by a closed lab.”
The release of Model Weights (numerical training to animate the system) will allow other companies and government agencies to deploy ML4 for their own purposes.
Mistral says its model is comparable in performance to China’s open-class competitor, the GLM-5.3, previously considered one of the best open-class models in the world, especially when it comes to cyberwork. However, Mistral’s model uses 47% fewer tokens per task, making it significantly more economical than GLM-5.3.
The company also claims that its model outperforms all other open models when it comes to agent coding tasks in the Coding Agent Index on AI benchmarking site Artificial Analysis.
“As cyber threats increase, ML4 brings cutting-edge sovereign cyber defense capabilities to businesses and governments around the world,” the company said Thursday. “Enterprises and public sector agencies can now download, deploy and own Frontier AI to detect, analyze and respond to threats across their code and IT infrastructure.”
ML4 weights are available via Hugging Face and its API.
Pierre Stock, Mistral’s vice president of science, told TechCrunch that exposing the model weights will give organizations more control over how they use AI to protect themselves.
“Specifically, this means that, for example, as a business or a government, you have access to cutting-edge intelligence to protect yourself from cyber-attacks. The model is open weight, so by definition access cannot be cut off,” he said.
Stock added that the model can be used to “identify, coordinate, and respond to critical security incidents in real time,” and that open sourcing allows companies to deploy the model on-premises within their own ecosystems.
Advances in AI cybersecurity have allowed security professionals to find bugs at a much higher rate than before, but given that “some malicious actors may actually try to exploit your technology,” Stock continued, it’s important for companies to “provide[the same]capabilities to everyone…to strengthen and strengthen the ecosystem.”
Separately, Stock cited concerns over recent incidents involving rogue AI agents, saying he recognized the risks but cautioned against overreacting.
“We don’t want to incite fear in people. We think this is a solvable problem. It’s a difficult problem to keep safe,” he said.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
