xiaomirobotics/xiaomi-robotics-1

This repository is the placeholder page for Xiaomi Robotics 1, a robot foundation model being built by Xiaomi's robotics team. According to the README, the model is a vision language action system, meaning it is trained to take in an image of what a robot sees along with a text instruction and produce the physical actions the robot should take to carry out that instruction. The headline claim is that the model was pre trained on more than 100,000 hours of real world robot trajectories, meaning actual recorded sequences of a robot moving and interacting with objects, rather than only simulated data. The authors describe this as a step toward scaling embodied intelligence, which refers to AI systems that act in the physical world through a robot body rather than only generating text or images. At the time this page was written, the actual code and the trained model weights had not been released yet. The README states they were expected to come out around July 16, 2026, and links out to a project website for more information, but the GitHub repository itself contains no source code, no installation instructions, and no usage examples. The only other content is an academic citation entry for researchers who want to reference the project in a paper. Because of this, there is nothing here yet that a developer, PM, or vibe coder can install or run. This repository is best understood as an announcement and citation placeholder for a robotics research project, not a working tool. Anyone interested in using it should check back later or visit the linked website for updates on when the code and model actually become available.