Mistral Launches Multimodal Neural Network

The French AI startup Mistral has unveiled its first multimodal model, capable of processing both images and text.

We dropped a new model — Pixtral 12B, our first-ever multimodal model. Enjoy! ?? https://t.co/uvXnpJf6mQ

— Sophia Yang, Ph.D. (@sophiamyang) September 11, 2024

Pixtral 12B is approximately 24 GB in size and boasts 12 billion parameters, a term used to describe its problem-solving capabilities. By comparison, the flagship Llama 3.1 405B has 405 billion parameters.

The AI is built on one of the Nemo 12B text neural networks and can answer questions about images.

Currently, the model can be downloaded from GitHub and Hugging Face, and it will later be available in the Mistral chatbot. This was announced by Mistral’s Head of Developer Relations, Sophia Yang.

We dropped a new model — Pixtral 12B, our first-ever multimodal model. Enjoy! ?? https://t.co/uvXnpJf6mQ

— Sophia Yang, Ph.D. (@sophiamyang) September 11, 2024

Mistral is considered a European rival to American AI giants like OpenAI. The firm raised €385 million in December 2023 and another €600 million in June 2024. Microsoft is among the company’s shareholders.

In May, the startup introduced Codestral, an artificial intelligence model for code generation.