More than 300 humanoid robots taking part in China’s second robot half marathon on Sunday will face tougher terrain designed to test their technical advances.
More than 70 teams, almost five times as many as last year, will compete in the 21-kilometer race in Beijing, which includes paved slopes and parks.
“It will certainly be interesting to see the progress in component durability and battery life compared to last year,” said Georg Stieler, managing director for Asia and head of robotics at Stieler, a technology consultancy.
Continues after advertising
“Manufacturers of humanoid robots need to find a balance between product quality, which is still constantly evolving, and price pressure.”
While all of last year’s participants were controlled remotely, this time nearly 40% of participating robots will travel the course autonomously, according to organizers, in a high-level demonstration of the industry’s growing capabilities.
In last year’s race, some robots crashed and fell near the starting line, while the winning model Tiangong Ultra, developed by the state-backed Beijing Innovation Center of Humanoid Robotics in collaboration with UBTech, finished in 2 hours and 40 minutes — comfortably ahead of its humanoid rivals, but more than twice the time of the human winner of the conventional race.
Tiangong Ultra will navigate “fully autonomously” this year, relying solely on its sensors to avoid obstacles, and will closely mimic human gait through large-scale data simulation training, the Center of Humanoid Robotics said in a statement.
“When the robot runs at speeds close to those of professional human athletes, the time window for perception and decision-making is extremely short, placing great demands on the computing power, algorithms and response speed of the system,” he said.
Videos on social media showing the robots training in Beijing at night this month showed some models successfully imitating human running and reaching speeds of 14 kilometers per hour, but the movements of others were more jerky, and some fell or crashed into railings, suggesting they may have difficulty reaching the finish line.
Continues after advertising
China dominates global installations of humanoid robots, accounting for more than 80% of the 16,000 units installed worldwide in 2025, according to Counterpoint Research. The top US supplier, Tesla, accounted for just 5% of global humanoid installations, according to the report.
Domestic market leaders AgiBot and Unitree each shipped more than 5,000 units last year — the most worldwide — while Unitree has committed to expanding production capacity to 75,000 humanoid robots per year.
“Dance disguised as work”
While the half marathon can be an entertaining spectacle, experts say the skills on display do not translate into the large-scale commercialization of humanoid robots in industrial settings, where manual dexterity, real-world awareness and capabilities beyond small-scale, repetitive tasks are crucial.
Continues after advertising
Currently, Unitree’s humanoid models are mainly used by research institutions, for dance performances and as interactive guides in service establishments, according to its initial public offering (IPO) prospectus.
And while some humanoids may be capable of completing a half marathon, even in China they are still years away from widespread domestic or industrial deployment, experts say.
“The reason our apps aren’t taking off is that the IQ of robots is very low. The models are bad, their success rates are low,” said Tang Wenbin, founder of embedded intelligence startup Yuanli Lingji, at a technology forum in Beijing last month.
Continues after advertising
“Honestly, the level of the entire industry is still at a very elementary stage… Right now, a lot of what we see is ‘dancing disguised as work’.”
The Chinese government has pointed to embodied intelligence, or physical AI, as one of the key sectors it wants to cultivate as it pursues automation to boost economic productivity and modernize traditional manufacturing.
Chinese robotics companies are still struggling to develop AI software that allows humanoids to match the efficiency of human workers in factories, while component makers face cost pressure, according to analysts.
Continues after advertising
As they seek to improve software, companies are investing resources in collecting large-scale real-world data, using human workers equipped with sensors and deploying more humanoids on the factory floor.
In 2024, UBTech has fewer than 10 humanoids in factories. Last year, that number jumped to more than 1,000.
This year, the company aims to launch 10,000 life-size humanoid robots, including new models adapted to a variety of commercial environments, said business director Michael Tam during a press visit to the company’s showroom in Shenzhen, southern China.
“When we talk about AI, it depends on the amount of data — especially high-quality data — that we can collect,” he said.
