The French AI startup Mistral has introduced its first generative AI models, designed to run on devices with limited computational resources, such as laptops and smartphones. These new neural network models, called Les Ministraux, can be used in various applications, ranging from text generation to more complex scenarios that require on-device data processing.

New AI Models for Edge Devices

The company launched two models — Ministral 3B and Ministral 8B. Both models can process context volumes of up to 128,000 tokens, allowing them to work with large datasets, such as texts up to 50 pages long.

“Our clients and partners are increasingly interested in local solutions for critical tasks, such as on-device translation, offline smart assistants, and local analytics. Les Ministraux provide powerful, low-latency computing solutions for these scenarios,” say the developers at Mistral.

Applications and Capabilities

The Les Ministraux models are designed for a wide range of applications, from basic text generation to complex tasks like autonomous robots or analytical systems that can function without a constant internet connection. This makes them attractive for those seeking enhanced privacy and reduced reliance on cloud computing.

Previously Released Products

In September 2024, Mistral introduced its first multimodal model, capable of working with both text and images. In July, the company released its flagship neural network, Large 2, which, according to the startup, is comparable in capabilities to models like GPT-4, Claude 3 Opus, and Llama 3.

Funding and Future Outlook

In June 2024, Mistral AI secured $640 million in funding, bringing the company’s valuation to $6 billion. Founded by former employees of Google DeepMind and Meta, Mistral positions itself as a strong European competitor to OpenAI.