Introducing “Les Ministraux”: Mistral's new AI models for efficient on-device computing

AI company Mistral has launched two new models, Ministral 3B and Ministral 8B, designed for on-device computing and edge use cases. Known as Les Ministraux, these models expand capabilities in the sub-10B category, excelling in knowledge, commonsense reasoning, function-calling, and efficiency. They support up to 128k context length, surpassing the current 32k on vLLM, with Ministral 8B featuring an interleaved sliding-window attention pattern for enhanced speed and memory efficiency.

Les Ministraux aim to deliver compute-efficient, low-latency solutions for privacy-focused applications such as on-device translation, internet-less smart assistants, local analytics, and autonomous robotics. In combination with larger language models like Mistral Large, they serve as efficient intermediaries for function-calling in multi-step workflows, capable of input parsing, task routing, and API calling across multiple contexts with minimal latency and cost. Both models are now available for use.