By NUPIAO News | Song Jianan
On August 19th, at the 2026 World Robot Conference, Xiaomi’s newest humanoid robot, fresh off a four-month internship at an automotive factory, was officially unveiled to the public.
This robot continues the “CyberOne” (Tieda) product line. It stands about 1.70 meters tall, weighs 66 kilograms, and boasts 66 degrees of freedom throughout its body, with 33 of those concentrated in the hands. Its hardware specs are tailored to match the physical build of frontline manufacturing workers. The exterior features a sleek, light-gray integrated design, with a matte silver helmet-style head. There are no unnecessary frills—it’s clearly built for industrial practicality.
At the expo, the robot took on the role of a “florist” and pulled off a series of interactive moves. It autonomously grabbed bouquets from the display, sprinkled scented sachets inside them, and then handed them to audience members. It also performed handshakes, fist bumps, and heart gestures. The finger movements, wrist rotations, and subtle body adjustments were impressively smooth, with no noticeable lag or stutter.
According to Xiaomi, none of these actions relied on pre-scripted programming. Instead, they were driven by an embodied AI large model that makes autonomous decisions in real time. The robot is still in the prototype development stage, and after its domestic debut, it’s set to travel to Europe for exhibition.
Unlike many products that are first publicly announced and then scurry to find real-world applications, this Xiaomi prototype had already completed production line training at Xiaomi’s EV factory before ever stepping foot on the expo stage.
Back in March, Xiaomi founder Lei Jun shared online that the company’s robots had started their “internship” at the EV plant, and that over the next five years, a large number of humanoid robots would be deployed to work in Xiaomi’s factories.
Lei Jun explained that Xiaomi built on its general-purpose VLA foundation model, Xiaomi-Robotics-0, and combined it with multimodal perception and reinforcement learning. This allowed the Xiaomi humanoid robot to initially achieve autonomous work in scenarios like self-tapping nut installation stations and bin handling. Key metrics like average time between failures and single-task success rates are steadily improving, and validation work around more classic workstations is ongoing.
Lei Jun pointed out that there’s a massive gap between lab work and real factory work—production tempo. In the lab, you can afford to fail ten thousand times to get one success. But in a real factory, you need to succeed ten thousand times out of ten thousand to meet production requirements.
According to Xiaomi’s released data, the robot ran autonomously for three straight hours at a self-tapping nut installation station in the die-casting workshop, achieving a 90.2% installation success rate while keeping up with the 76-second production line cycle. Xiaomi describes this as the first step toward stable application of its humanoid robots in smart manufacturing.
Then in July, Xiaomi dropped a video of the robot in action. The footage was shot in the logistics area of the general assembly workshop at Xiaomi’s EV plant. In the sequence showing center console side cover sorting, the robot used its full-body multi-degree-of-freedom coordination, leaning in slightly as its dexterous hands reached into the bin to grab the cover. After retrieving the part, it performed a hand-to-hand transfer to adjust its grip in real time. Once adjusted, the robot lifted the flexible cover with both hands and placed it into the corresponding slot on the rack cart, completing the flexible part sorting process.
In the folding and recycling task, the robot faced empty plastic logistics bins. It used its dexterous fingers to precisely pop open the bin’s locking rings, then coordinated both arms to fold, press, and snap the bin shut, compressing the open bin into a flat stack. After folding, it pushed the collapsed bins into their designated storage spot, completing the entire material recycling workflow.
Lei Jun noted that by this point, the success rate for the bilateral operation at the self-tapping nut installation station had jumped from 90.2% to 98%. The success rates for center console side cover installation and bin folding both hit 90%, and the robot achieved long-duration operation with flexible workpieces for the first time.
Xiaomi first started exploring humanoid robots around 2021, and officially launched the first-generation CyberOne (Tieda) robot the following year. At that time, the product was more about validating the technology. The first-gen robot had 21 degrees of freedom and focused on interactive capabilities—it was still a long way off from industrial-grade performance.
In terms of product positioning, Xiaomi’s humanoid robot isn’t chasing standalone hardware sales. Instead, it’s designed to support Xiaomi’s “Human x Car x Home” full-ecosystem strategy. The tech route uses a full-size bipedal design with a hybrid “brain + cerebellum” architecture. The company has also developed its own Mi-Sense depth vision perception system, is pushing forward with embodied AI large models, and has open-sourced the Xiaomi-Robotics-0 robot model to lower the barrier for industry adoption.
Lei Jun mentioned in July that Xiaomi’s embodied foundation model, Xiaomi-Robotics-1, ranked first globally in both the Robocasa365 and RoboDojo embodied intelligence benchmarks. The Xiaomi robotics team has basically completed its “hardware, data, model” layout, and the next step is to push robots into more open and complex real-world environments.
Just a day before the expo, on August 18th, Xiaomi released its Q2 2026 financial results. Total revenue came in at 108.9 billion yuan, down 6.1% year-over-year but up 9.9% quarter-over-quarter. Adjusted net profit was 6.2 billion yuan, a 42.6% decline year-over-year but a 2.4% increase from the previous quarter.
Despite the dip in revenue and profit, Xiaomi’s R&D spending hit 9.23 billion yuan in Q2, up 18.9% year-over-year, with heavy investment going into smart EVs, AI, and other core areas.
Lei Jun has previously revealed that Xiaomi’s cumulative R&D investment over the past five years exceeded 100 billion yuan, and the plan is to surpass 200 billion yuan over the next five years. Over the next three years, Xiaomi plans to invest more than 60 billion yuan in AI. This year alone, AI-related R&D and capital expenditure will exceed 16 billion yuan. Humanoid embodied intelligence is also one of its key focus areas.
That said, Xiaomi’s robotics business is still in the prototype iteration and internal validation phase. It doesn’t yet generate independent external revenue. For now, its value lies mainly in reducing factory costs and accumulating technical know-how. There are still plenty of engineering challenges to overcome before it can move toward large-scale external commercialization.
