Xiaomi robotics achieves 8-hour factory shifts and AI tasks

Xiaomi robotics achieved major operational milestones across industrial and domestic settings, validating an 8-hour continuous manufacturing shift alongside a 30-minute multi-room home trial. Designed for automated assembly lines and smart household management, the humanoid system's primary constraint remains its staged trial deployment rather than broad commercial availability.

Industrial Line Integration: 8-Hour Autonomous Operation

During the 2026 World Internet Conference, where Xiaomi received the Outstanding Contribution Award for innovative breakthroughs and its robotics division earned the Leading Science and Technology Award, the company unveiled new progress across its primary testing grounds. In factory environments, Xiaomi's humanoid robot completed an 8-hour work task at a center console side cover sorting station.

The test satisfied full-day production requirements for the workstation. Throughout the operation, the robot demonstrated autonomous error correction and production-line system coordination, establishing initial verification for factory-grade manufacturing capabilities.

Domestic Automation and the Xiaomi-Robotics-1 Foundation Model

In residential settings, Xiaomi deployed its embodied foundation model for trial operations at the Xiaomi Youth Apartments. Powered by the Xiaomi-Robotics-1 pre-trained model within an agent framework, the robot executed continuous multi-room operations lasting over 30 minutes, spanning three distinct rooms and addressing more than 50 individual sub-tasks.

The underlying software relies on two specialized models developed by Xiaomi's engineering teams:

• Xiaomi-Robotics-0: A vision-language-action (VLA) model utilizing a hybrid brain architecture that attains an 80-millisecond inference latency on consumer-grade graphics cards.
• Xiaomi-Robotics-1: An embodied foundation model pre-trained on 100,000 hours of real-world operational trajectories, achieving the top composite ranking on both the RoboCasa 365 and RoboDojo benchmarks in July 2026.

Using Xiaomi-Robotics-1 alongside iterative post-training data, the team created a closed-loop agent pipeline linking data collection, model inference, and real-time execution. The domestic workflow autonomously maps and resolves sequential tasks, including tool preparation, sofa and tabletop tidying, trash disposal, object sorting, room inspections, full-home re-checks, and final exit procedures.

As humanoid platforms expand from repetitive industrial assembly into complex, multi-room living environments, which setting presents the steeper hurdle for real-world reliability: the strict precision tolerances of factory lines or the unpredictable chaos of household clutter?