French AI company Mistral has released the weights for Mistral Large 4, its open-weight AI model with one trillion parameters. The move lets businesses and governments download the model and run it on their own systems.
The model, known as ML4, carries the nickname “Le Chonk.” Mistral first introduced it earlier this month.
Weights are the numbers learned during training that make an AI model work. By releasing them, Mistral allows other companies and government agencies to deploy ML4 for their own needs.
Mistral Targets Cyber Defense
Mistral is promoting the model as a tool for cybersecurity. The company said ML4 gives enterprises and governments “state-of-the-art sovereign cyber defense capabilities” around the world.
“Enterprise and public sector institutions can now download, deploy, and own frontier AI to detect, analyze, and respond to threats across code and IT infrastructure,” the company said Thursday.
Pierre Stock, Mistral’s Vice President of Science, told TechCrunch that open weights give organizations more control. He said the model offers access to advanced AI for defense that “can’t be cut off by definition because the model is open-weighted.”
Stock said the model could help “identify, coordinate, and respond in real time to major security incidents.” He added that companies can run it on-premises inside their own systems.
He said AI tools are helping security workers find bugs much faster than before. He argued this ability should reach everyone quickly, since “some malicious actors might try to actually leverage their own technology to attack.”
How ML4 Compares to Rival Models
Mistral says ML4 performs on par with GLM-5.3, a Chinese open-weight model. GLM-5.3 had been seen as one of the best open-weight models, especially for cyber work.
According to Mistral, ML4 uses 47 percent fewer tokens per task than GLM-5.3. That makes it cheaper to run for the same work.
The company also claims ML4 ranks above all other open models for agentic coding tasks. That ranking comes from the Coding Agent Index run by benchmarking site Artificial Analysis.
Mistral says it reached “state-of-the-art performance” using a “fraction of the compute and capital committed by closed labs.” CEO Arthur Mensch has said AI should “be a tool for empowerment, not dominance.”
The release comes as open-weight developers like Mistral compete with closed-model companies such as OpenAI and Anthropic. Open-weight firms argue that businesses and governments should not have to depend on outside AI providers for advanced tools.
Stock also spoke about recent incidents involving rogue AI agents. He acknowledged the risks but warned against overreacting.
“We don’t want to fearmonger people. We think it’s a problem that is solvable. It’s a hard problem to have good security,” he said.
The ML4 weights are now available to download on Hugging Face. The model can also be accessed through Mistral’s API.
