tl;dr: Xiaomi-Robotics-1 is a robot foundation model pre-trained on 100,000 hours of embodiment-free UMI trajectories, then post-trained with ~7,200 hours of real-robot data to align it with physical embodiments and natural-language instructions. The team reports clean scaling behavior—larger models and more pre-training data yield predictably better real-robot success rates—and achieves 75-85% success on new tasks with under 10-40 hours of demonstrations, roughly doubling the π0.5 baseline while topping four simulation benchmarks.