• 2 min read
Xiaomi says Robotics-1 learned from 100,000 real-world hours
Xiaomi has unveiled Robotics-1, its first foundation model for embodied AI and robotics, trained on 100,000 hours of real-world data.
Image: ITzine
Xiaomi has introduced Xiaomi-Robotics-1, its first foundation model for embodied AI and robotics. The company says the system was trained on 100,000 hours of real-world data and is designed to work across different types of robots rather than a single lab setup.
According to Xiaomi, the model is built to handle mobile control and adapt to new conditions without lengthy platform-by-platform tuning. The company first pre-trained it on large datasets, then fine-tuned it on several robotic platforms.
In testing, Xiaomi claims first place in RoboCasa365 and RoboDojo. It is also making a notable open-source play: the company says it will release both the model’s source code and weights, giving developers access to more than just a demo.
Recommended reading
Hassabis says STEM makes you 10x better at AI
Xiaomi is not the first to enter this field. In 2023, Google DeepMind introduced RT-2, a model designed to transfer knowledge from the internet into robotic actions. In 2024, NVIDIA announced Project GR00T for humanoid robots, and researchers from Stanford, Berkeley, and Google DeepMind released the open OpenVLA.
Xiaomi’s pitch is different, at least for now: real-world data and openness. If the company publishes the weights soon as promised, robotics developers will have another significant open-source tool to work with.
AI Editor
Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.
via ITzine
