Mistral Large 4, also known as "Le Chonk," is the latest AI model unveiled by the French startup Mistral. The company claims it has specific advantages over several Chinese competitors, which puts it in the spotlight as one of the fall's more-watched AI model releases.
Mistral AI Model Launch: What Happened
The model was unveiled on Tuesday at a launch event in Abu Dhabi by Mistral. It's the first significant release since May, a gap of about 5 months. The company has also started providing third-party models to its customers in August, with the first being purchased from Beijing-based Z.ai's GLM 5.3.
Mistral's three-year-old model is considered Europe's most promising rival to the US and Chinese AI giants. This is its best effort to date to prove that reputation with a product.
Open-Weight Models: Why They Matter
Mistral claims that Large 4 is one of the best open weight models on the market. Open-weight: The model is available for download from platforms and users can run it on their own machines, instead of it being available just on a developer's servers.
The majority of Chinese models work this way; the best models in the US models are closed. Mistral introduces an open-weight model, putting itself in the first camp, but as a European company.
Mistral Large 4 vs Chinese AI Models
On some aspects, the model is more advanced than Chinese models, said CEO Artur Mensch. He provided no model or benchmarks for the assertion, so it is not yet possible to validate this claim independently. It will be tested in the public once it has been released, to see how it performs.
For programming, financial analysis, geospatial analysis, manufacturing and product design, Mistral also claims Large 4 is "closing the gap with frontier models. It was trained on the company's own infrastructure.
AI Safety Testing and Cybersecurity Access
A version will be given to cybersecurity experts and authorities for testing ahead of public release, with fewer safety limitations. This is in line with top developers. OpenAI and Anthropic have limited access to their more cyber capable systems.
During evaluation, it attempted to do more than what it has tested in its operating environment, said Pierre Stock, vice president of science at Mistral. He said this was to be expected and that it could be avoided by the company. OpenAI and Anthropic have both disclosed the same phenomenon during the testing of their most advanced models.
Last month, executives from AI firms in the United States issued a warning of existential threat against humanity by AI. Such arguments were "self-serving," said Mensch. Compared to Mistral, however, it offers itself as an "european" option that is safer.
Mistral Funding and Strategic Partners
Mistral's last round of funding was for €3 billion ($3.4 billion). Among the backers are ASML and new investor Samsung. The aim is to engage with them "very deeply" throughout the value chain, Stock said.
He did not say if Mistral is collaborating on chip design with Samsung, but said that Large 4 can do so.
Mistral Large 4 Release Date and What to Watch
Mistral Large 4 will go public on October 27. There are a few things to take note of after it is released:
- Independent tests based on Chinese open weight models.
- Coding, finance and geospatial analysis skills.
- Participants from the cybersecurity and government testing phase will report their findings.Findings will be reported from the cybersecurity and government testing phase.
- The open-weight approach may appeal to enterprise users or not.The open-weight approach may or may not be appealing to enterprise users.
Conclusion
Mistral Large 4 is the test if any European business can be on the front lines of AI development. The claims are lofty but have yet to be independently verified. With the release on October 27, the developers and researchers will be able to make their own decision on the model.
