Mistral ships Robostral Navigate — an 8B embodied navigation model that moves robots through unseen environments using one RGB camera
Robostral Navigate is an 8-billion-parameter embodied navigation model that lets a robot traverse environments it has never seen using nothing but a single RGB camera. No depth sensor, no lidar, no prior map.
The sensor constraint is the whole point. Navigation stacks have historically bought reliability with hardware — lidar, depth cameras, wheel odometry, prior maps. Doing it from one RGB stream at 8B parameters puts capable navigation inside the cost and power envelope of ordinary hardware rather than research platforms.
It also fits the convergence visible in FLUX 3's image-to-video-to-action framing: vision models are increasingly producing actions rather than descriptions, and the architectures are starting to look like each other.
The Neuron — Everything That Happened in AI Today → · AI Release Tracker — Latest AI Model Releases — July 2026 →