Mistral, the French AI startup, has launched Mistral Large 4, a new AI model that the company claims outperforms many rival open-weight models, including certain Chinese models. The model is set to be publicly available on October 27, with a version for cybersecurity experts and government authorities to test beforehand.
TL;DR
- Mistral Large 4 is claimed to outperform many rival open-weight models, including certain Chinese models, with a focus on cybersecurity.
- The model will be publicly available on October 27, with a version for cybersecurity experts and government authorities to test beforehand.
- Mistral has positioned itself as a safer, European alternative to US AI companies, raising €3 billion in its latest funding round.
What happened
Mistral, a French AI startup, unveiled Mistral Large 4, a new AI model that the company claims outperforms many rival open-weight models, including certain Chinese models. The model is set to be publicly available on October 27, with a version for cybersecurity experts and government authorities to test beforehand. This is Mistral's first major release in about five months, with the last one being in May.
CEO Artur Mensch, speaking at a launch event in Abu Dhabi, claimed that Mistral Large 4 outperforms Chinese models in certain aspects, including cybersecurity. However, he did not specify which Chinese models or benchmarks he was referring to. The model, officially named Mistral Large 4 but nicknamed Le Chonk, will be made publicly available on October 27. Before the release, cybersecurity experts and government authorities will be given access to a version with fewer safety restrictions to test its capabilities, the company said.
Why it matters
Mistral's claim that its new model outperforms certain Chinese models in cybersecurity is significant, as it challenges the narrative that European AI companies cannot compete with their Chinese counterparts. This could have implications for the global AI market, as Mistral positions itself as a strong European challenger to US and Chinese competitors.
The model's focus on cybersecurity is particularly relevant, as AI models with advanced cyber capabilities have raised concerns about potential safety risks. Mistral's claim that it has been able to prevent its model from going beyond its testing environment is a significant development in this regard. However, the company's dismissal of arguments that AI could pose an existential threat to humanity may be seen as controversial, given the recent warnings from US AI companies.
Key facts
- Mistral Large 4 is claimed to outperform many rival open-weight models, including certain Chinese models, with a focus on cybersecurity.
- The model will be publicly available on October 27, with a version for cybersecurity experts and government authorities to test beforehand.
- Mistral has positioned itself as a safer, European alternative to US AI companies, raising €3 billion in its latest funding round.
- Mistral Large 4 is claimed to rank among the strongest open-weight systems available, according to Pierre Stock, Mistral's vice president of science.
- The model is claimed to be capable of designing computer chips, but Mistral declined to comment on whether it is currently working with Samsung on chip development.
- Mistral has partnered with backers including ASML and new-investor Samsung across the entire value chain, according to Pierre Stock.
Context
Mistral's launch of Mistral Large 4 is significant in the context of the global AI market, where US and Chinese companies have traditionally dominated. The company's claim that its model outperforms certain Chinese models in cybersecurity is a direct challenge to this dominance, and could have implications for the future of the AI industry.
The model's focus on cybersecurity is also relevant in the context of recent concerns about the potential safety risks of advanced AI models. Mistral's claim that it has been able to prevent its model from going beyond its testing environment is a significant development in this regard, and could have implications for the future of AI safety.
