Mistral's new 1T model aims to leapfrog closed and open rivals
Key Takeaways
- Mistral AI released Mistral Large 4, a one-trillion-parameter multimodal model.
- The model is positioned as a sovereign European alternative to US and Chinese AI systems.
- Open weights will be released in three weeks after thorough safety testing.
- Training utilized only 4,000 NVIDIA GPUs, showcasing high computational efficiency.
French artificial intelligence laboratory Mistral AI has launched Mistral Large 4, a massive multimodal model with one trillion parameters designed to compete with leading American and Chinese AI systems. Positioned by French leadership as a sovereign European alternative, the model aims to bridge the gap between restrictive closed models and open-weight alternatives.
Nicknamed "Le Chonk," the model is initially available through a guarded endpoint, with open weights expected in three weeks after safety checks. Trained efficiently on just 4,000 NVIDIA GPUs, Mistral Large 4 targets enterprise-critical domains such as cybersecurity, finance, and semiconductor chip design, supported by major investors like ASML and Samsung.
Recommended for you
Tools and services we trust to boost productivity and content workflows.
Browse picks