Mistral's 8B Robostral Navigate hits 76.6% on R2R-CE using only one RGB camera
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
Robostral Navigate achieves 76.6% success on unseen R2R-CE benchmarks using only a single RGB camera, outperforming multi-sensor approaches by 4.5 points while being more efficient. This enables cost-effective deployment of autonomous navigation in real-world environments like offices and warehouses without expensive LiDAR or depth sensors, simplifying hardware requirements and scaling robotics applications.
Robostral Navigate is an 8B embodied AI model that achieves a 76.6% success rate on the unseen R2R-CE benchmark using only a single, standard RGB camera with zero depth sensors or LiDAR, outperforming the leading multi-sensor models by 4.5 points. For production engineering, this eliminates the need for expensive, heavy sensor suites and complex hardware calibration, allowing you to deploy reliable, long-horizon language-to-motion navigation on cheap, lightweight robots like basic wheeled or flying platforms. By predicting target coordinates directly within the 2D camera view instead of absolute metric spaces, the model remains robust to variations in camera hardware and scale, vastly simplifying fleets built with heterogeneous hardware.