French AI company Mistral has released Mistral Large 4—nicknamed Le Chonk—a new 1 trillion-parameter AI model that the company claims is “by far the most capable open-weight model developed outside of China,” according to Wired.
The model is currently available in preview through a public guardrail endpoint, with plans to release its full weights by the end of October after safety testing is complete, according to TechCrunch. “In the meantime, we’ll work with trusted partners and governments to make sure that the open source weights can be used to defend, but not to [perform] malicious attacks,” Mistral VP Science Pierre Stock told TechCrunch.
According to Wired, Mistral cofounder and chief scientist Guillaume Lample said the model is “very, very close” to some proprietary models and was trained from scratch, distinguishing it from Chinese labs that the US government has accused of using distillation techniques. TechCrunch reports that the model was trained using only 4,000 Nvidia GPUs, “which is two to three times less than our Chinese competitors, and significantly less than the closed source competitors,” according to Stock.
The model is optimized for coding, cyberdefense, manufacturing, finance, and electrical engineering, according to Wired. Mistral positions Le Chonk as what French President Macron described as “a third way in AI,” offering an alternative to both closed American models and open Chinese models, TechCrunch reports.