Xiaomi debuted its new-generation humanoid robot on August 19 at the 2026 World Robot Conference, following four months of production-line training inside one of the company's auto factories1. The prototype, continuing the Tieda product line, stands 1.70 meters tall, weighs 66 kilograms, and has 66 degrees of freedom — 33 of them in the hands. At the conference, the robot performed as a florist, autonomously grasping bouquets and handing them to visitors, completing handshakes and fist bumps. These actions relied on autonomous decision-making driven by models rather than preset scripts.
The robot draws on the open-source Xiaomi-Robotics-0 foundation model, which combines multimodal perception and reinforcement learning. It uses a brain-plus-cerebellum hybrid architecture and Xiaomi's self-developed Mi-Sense depth vision system. In factory trials, the humanoid achieved autonomous work at self-tapping nut stations and material-box handling. At a die-casting workshop nut station, it ran autonomously for three hours with a 90.2% installation success rate, meeting the 76-second cycle time. Founder Lei Jun said success at the nut station later rose from 90.2% to 98%, and that cover loading and box folding reached 90%. Lei Jun said long-duration flexible-part work was achieved for the first time.
In July, Xiaomi released footage of additional tasks: at a center-console side-cover sorting station, the robot reached into a material box, switched hands to adjust its grip, and placed the flexible cover into the corresponding slot. In box folding, it used a dexterous hand to open a lock ring, coordinated both arms, and pushed folded boxes to a storage spot. In March, Lei Jun said robots had entered the factory for "internships" and that large numbers would eventually work in Xiaomi factories. The prototype will head to Europe after its conference debut.
Xiaomi entered humanoid robotics around 2021 and released the first-generation CyberOne Tieda the following year, which had 21 whole-body degrees of freedom and focused on interaction. The robot business remains at the prototype-iteration stage with no independent external revenue; its value is currently in factory cost reduction and technology accumulation2.
Separately, Xiaomi also released two embodied foundation models. In July, Lei Jun said the Xiaomi-Robotics-1 model ranked first globally on the Robocasa365 and RoboDojo benchmarks.
The robotics push sits within a broader AI investment surge. Xiaomi reported Q2 2026 revenue of 108.92 billion yuan on August 18, with adjusted net profit of 6.22 billion yuan and attributable net profit of 9.46 billion yuan, doubling quarter over quarter. Q2 R&D spending reached 9.2 billion yuan, up 18.9%, directed mainly at AI and chips. Lei Jun called AI the next decade's largest opportunity, with 2026 AI R&D and capital expenditure exceeding 16 billion yuan and 60 billion yuan planned over three years.
On the AI model front, Xiaomi released MiMo-V2.5 in Q2, which topped OpenRouter's monthly and weekly call charts over DeepSeek-V4 Flash. CFO Lin Shiwei said the result rests on advantages in model capability, inference efficiency, and cost control, and a new MiMo model is training for imminent release. Lei Jun also declared a 50 billion yuan, 10-year push to build Xuanjie chips.
ANALYSIS The jump from 21 to 66 degrees of freedom between the CyberOne and the new Tieda prototype, combined with the shift from scripted interaction to model-driven autonomous factory tasks, marks a substantial generational leap in Xiaomi's robotics hardware and software stack. Deploying the humanoid on its own auto production lines gives Xiaomi a closed loop — real factory data feeding back into its embodied models — that pure robotics startups without manufacturing operations cannot easily replicate.