Companies
Mistral Launches Large 4 AI Model, Targets US Market with Competitive Features
Mistral AI has unveiled its flagship AI model, Mistral Large 4, also known as “Le Chonk,” on October 6, 2026. The French company positions this model as the strongest open-weight AI system developed outside of China, asserting that it outperforms certain Chinese rivals. The model is scheduled for public release on October 27, allowing users to evaluate its performance.
Mistral Large 4 (ML4) boasts 1 trillion parameters, though only 49 billion are active at any given moment, thanks to its Mixture-of-Experts (MoE) design. This approach allows the model to efficiently route requests to a subset of its internal “experts,” akin to directing patients to specialized doctors in a hospital. Additionally, ML4 is natively multimodal, built to handle various data types beyond just text.
The training of ML4 took approximately two months, utilizing around 4,000 NVIDIA Grace Blackwell GPUs in Mistral’s European data centers. According to Mistral, ML4 achieved a score of 38 on the Artificial Analysis Intelligence Index, the highest for any non-Chinese open-weight model, although it still trails behind the Chinese model MiMo-V2.6-Pro, which scored 46.
Mistral is targeting ML4 for specific applications, including cybersecurity, finance, coding, manufacturing, and visual grounding, supporting over 160 languages, including all official EU languages. Currently, access to ML4 is limited to select developers, but pricing details have been released, with input costs at $1.36 per million tokens and output at $4.18 per million tokens.
FAQ
What is the Mistral Large 4 AI model?
Mistral Large 4, also known as 'Le Chonk,' is a flagship AI model developed by Mistral AI, featuring 1 trillion parameters and designed to outperform certain Chinese rivals. It utilizes a Mixture-of-Experts (MoE) architecture for efficient processing.
When will Mistral Large 4 be publicly released?
Mistral Large 4 is scheduled for public release on October 27, 2026, allowing users to evaluate its performance.
What are the main applications for Mistral Large 4?
Mistral Large 4 is targeted for applications in cybersecurity, finance, coding, manufacturing, and visual grounding, supporting over 160 languages, including all official EU languages.
How does the Mixture-of-Experts design work in ML4?
The Mixture-of-Experts design allows Mistral Large 4 to efficiently route requests to a subset of its internal 'experts,' optimizing performance by directing tasks to specialized components, similar to how patients are referred to specialized doctors.
What are the pricing details for using Mistral Large 4?
The pricing for Mistral Large 4 includes input costs of $1.36 per million tokens and output costs of $4.18 per million tokens.