ZDNET’s key takeaways
- Mistral’s open weight ML4 model Le Chonk is in preview.
- ML4 was trained with fewer GPUs than OpenAI’s Astra, but competes.
- Open models are positioned as democratic defenders of AI attacks.
French AI lab Mistral has shipped its latest model, Mistral Large 4 (ML4) – and it is positioned as the open solution for all your defense requirements.
Also: Open weights vs. closed: An AI civil war is underway, and the stakes are existential
One result of recent AI security incidents is pitting open and proprietary models against each other. Initially considered less secure due to their malleability, open models are seen as a new option for cyberdefense after proprietary models from Anthropic and OpenAI proved just as risky.
Mistral said ML4, which the company dubbed “Le Chonk” for its trillion-parameter size, is built for security that remains under user control — unlike proprietary models to which border labs can technically revoke access at any time.
“The cyber defense capabilities allow enterprises and governments to defend themselves against threat actors who jailbreak closed models to carry out cyber attacks,” said Mistral co-founder Guillaume Lample.
Le Chonk and security
After a hack-filled summer that put AI model security under the spotlight, everyone is looking for a reliable AI security solution that meets their needs. ML4 prioritizes cyber defense capabilities, customizable advertising control and data sovereignty.
Also: This new ChatGPT scam tricks you into installing malware – how to spot the trap
In a briefing, Lample and Mistral’s VP of Science Pierre Stock emphasized that security is a key requirement for enterprise customers, which echoes a further industry trend. After the Hugging Face breach, Mistral was one of many companies that signed Nvidia’s Open Secure AI Alliance, a cross-industry partnership that argues that open models are crucial to democratizing the defense against increasingly common AI security incidents.
“ML4 is the beginning of a leading generation of open-weight, customizable, cybersecurity models that enterprises can fully own and control, without vendor lock-in,” Mistral wrote. “Companies and states must not rely on a closed model vendor that could arbitrarily disable their cyber defense capabilities.”
By the logic of Nvidia, and its alliance that aims to democratize AI security tools, the race is between Mistral and other open models to achieve state-of-the-art security capabilities.
“In absolute terms on cyber capabilities, ML4 surpasses the best models of Kimi, Deepseek and Meta,” a Mistral spokesperson told ZDNET via email.
Also: Who owns AI risk at work? Business and tech leaders can’t agree, PwC survey finds
Le Chonk is now available in public preview. Mistral said it will release the model weights on October 27. That time gap gives the lab a month “to work with developers, cybersecurity leaders and government authorities to further assess the capabilities and behavior of ML4 in real-world environments” – a practice common to proprietary American labs such as OpenAI, Google, and Anthropies over concerns about model capabilities.
Similar to Anthropic’s Project Glasswing and OpenAI’s rollout of Astra, initial test partners will get access to a less protected version of ML4 with “enhanced cybersecurity capabilities.”
Outside of security, Mistral said Le Chonk excels in finance and multimodal use cases. The company said it is still waiting for final benchmarks. However, early third-party analysis shows ML4 competing on par with pricier proprietary models like GPT-6 Astra in certain computer vision tasks (as in the benchmarks below), as well as impressive open-weight Chinese models like Kimi K3. Le Chonk met or slightly outperformed DeepSeek models on financial work assignments, achieving a new high of 15% for open-weight models on the Harvey Legal Agent benchmark.
As a reminder, benchmark scores themselves should be taken with a grain of salt, especially considering how many models cheat.
Also: The AI models that cheat the most, according to new CAIS benchmark
Chinese labs like DeepSeek and Moonshot (which developed Kimi models) have been accused of distilling or ripping off proprietary models from American labs to gain their competitive advantage. Mistral reiterated that it is not participating in that process.
“We are completely separate from other models, and we do not take inspiration from them,” Stock said in the briefing.
The company also emphasized its commitment to sovereignty, an equally hot topic, especially in Europe.
“Customers will soon have flexible deployment options: self-deployment or access via our API in the region of their choice, including our European sovereign region where data remains under EU jurisdiction,” Mistral wrote.
More for less computing
Training a competitive model in a computer shortage is no small task for a trimmer lab like Mistral, which does not have the same resources as a pre-IPO giant like Anthropic.
“ML4 was trained from scratch on 4,000 Nvidia Grace Blackwell GPUs over two months, deployed in Mistral’s own data centers in Europe,” the company said, adding that the previews also run on the same GPUs. For context, Nvidia CEO Jensen Huang told X that OpenAI trained GPT-6 Astra on about 100,000 GPUs. That’s quite the faction.
“We expect the model to be significantly improved in the coming months. This model will also serve as the basis for a new wave of specialized and optimized models of Mistral,” the company added.
See full bio