Mistral released an updated version of its Large 4 artificial intelligence model on Thursday, Oct. 29, highlighting enhanced coding and cybersecurity features. The French startup stated that the new release builds upon a public preview launched earlier in the month.
The company noted that the response to the initial preview exceeded expectations. In its announcement, Mistral thanked its developer community, early-access customers, and partners for their support and feedback during the testing phase.
Cybersecurity Performance and Refusal Rates
According to the announcement, the model demonstrates strong performance on cybersecurity benchmarks. Mistral reported that the system achieves an average refusal rate of 96.6% on malicious cyber requests. The company claimed this is the highest refusal rate it measures among open models.
The model, nicknamed “Le Chonk” and featuring an orange cat logo, is designed for companies to run on their own systems. This approach aims to reduce reliance on third-party AI models. The nickname “chonk” refers to internet slang for a hefty cat.
Open-Weight Strategy and Industry Context
Mistral Vice President of Science Pierre Stock emphasized the importance of open weights in a Thursday Reuters report. He stated, “I’m not comfortable living in a world in which open weights do not exist.” Stock added that once the model is released, it is impossible to contain access, which protects users from being cut off but also means anyone can use it.
The release occurs amid an ongoing debate in the AI sector regarding open versus closed-source models. This discussion has divided the industry, with financial considerations playing a significant role for middle-market companies. The question for finance chiefs is less about ideology and more about whether the savings, flexibility, and control offered by open models justify the responsibility for underlying infrastructure.
As AI moves beyond isolated chatbots into finance, procurement, treasury, compliance, and enterprise software, this trade-off is becoming more important. Mistral described Large 4 as the best open-weight model for coding and one of the world’s most capable open models overall.
Source: PYMNTS
