Mistral's new 1T model aims to leapfrog closed and open rivals
French AI lab Mistral AI has released Mistral Large 4, a new large multimodal model aiming to leapfrog both American and Chinese rivals.
TL;DR
- Mistral AI has released Mistral Large 4 (ML4), a new large multimodal model.
- ML4 aims to be a "third way in AI," offering an alternative to closed and open-weight models.
- The model, nicknamed Le Chonk, has 1 trillion parameters and is not yet an open-weight model.
- Mistral plans to release ML4's weights in three weeks after safety testing, focusing on defense applications.
- ML4 was trained using 4,000 Nvidia GPUs, significantly less than competitors.
- Optimized use cases include cybersecurity, finance, and chip design, supported by backers like ASML and Samsung.