Mistral AI

Summary

About

About Mistral AI

Mistral AI develops language, vision and reasoning models alongside enterprise tools for deploying them. Its robotics programme includes Robostral Navigate, an eight-billion-parameter vision-language-action model that converts a single RGB camera view and a natural-language destination into local movement instructions.

Robostral Navigate can guide wheeled, legged or flying platforms without lidar or a depth camera, predicting either a target image position and heading or a short local displacement. Training combines 2.4 million simulated trajectories across 350,000 scenes with reinforcement learning, and the model is designed to generalise across robot size and previously unseen environments. Mistral released the model in July 2026 for embodied-navigation development.

PRICE
Range 1Y
Return
High
Low
Updated

Timeline

Events

Relations

Relationships

  • Helsing

    AI Development Partner - Helsing and Mistral AI are jointly developing defence Vision-Language-Action models.