China Humanoid Robot Half-Marathon to Showcase Technical Leaps
China is set to showcase its robotics ambitions as more than 300 humanoid robots from over 70 teams compete in Beijing’s second robot half-marathon, a 21-kilometre race designed to test endurance, balance, autonomy and real-world decision-making. Nearly 40% of the robots will run fully autonomously up significantly from last year when all robots were remotely controlled while last year’s winner, Tiangong Ultra, returns with full sensor-based navigation. The event highlights China’s dominance in humanoid robotics, with the country accounting for over 80% of global installations in 2025 and companies like Unitree and UBTech rapidly scaling production. However, experts warn that while the marathon demonstrates technical progress, most humanoid robots are still far from widespread industrial or home deployment, with reliability, dexterity and real-world intelligence remaining major challenges.
China is preparing for a major display of robotics innovation as more than 300 humanoid robots are set to participate in the country’s second robot half-marathon on Sunday in Beijing. The event is designed to test how far humanoid robotics has advanced while highlighting China’s broader ambition to make robotics and embodied AI a key pillar of its economy.
More than 70 teams will compete in the 21-kilometre race, nearly five times more than last year’s event. Unlike the smoother route used previously, this year’s course includes paved slopes, parkland and more challenging terrain aimed at pushing the robots’ endurance, balance and decision-making systems to the limit. Reuters reported that nearly 40 percent of the participating robots will run autonomously, a major leap from last year when all robots were remotely controlled.
Industry experts say the event offers a clear benchmark for how quickly humanoid robotics is evolving. Georg Stieler, Asia managing director and head of robotics at Stieler, said it will be especially important to observe improvements in battery life and the durability of critical components compared to last year’s race. He noted that manufacturers are under pressure to balance product quality while keeping costs manageable as the technology continues to evolve rapidly.
One of the most closely watched participants will be Tiangong Ultra, the humanoid robot that won last year’s race. Developed by the state-backed Beijing Innovation Center of Humanoid Robotics in collaboration with UBTech Robotics, the robot finished the previous marathon in 2 hours and 40 minutes, more than twice the time of the human winner. This year, Tiangong Ultra will run fully autonomously, relying entirely on onboard sensors to avoid obstacles and mimic human running patterns through large-scale simulation training.
According to the robotics center, running at speeds close to professional human athletes requires extremely fast perception and decision-making. This places heavy demands on computing power, algorithms and system response speed. Social media videos from recent night training sessions in Beijing showed mixed results. Some robots successfully reached speeds of up to 14 km per hour and closely resembled human movement, while others stumbled, fell, or even crashed into barriers, showing that reliability remains a major challenge.
China currently dominates global humanoid robot installations, accounting for more than 80 percent of the world’s 16,000 installed units in 2025, according to Counterpoint Research. Domestic leaders such as AgiBot and Unitree Robotics each shipped more than 5,000 units last year, the highest globally. Unitree has also announced plans to expand its annual humanoid robot production capacity to 75,000 units. In comparison, Tesla accounted for only 5 percent of global humanoid installations.
Despite the excitement around the half-marathon, experts caution that success in a race does not necessarily translate into real-world industrial usefulness. Humanoid robots still struggle with manual dexterity, real-world perception and tasks that go beyond repetitive motions in controlled environments. Many of today’s robots are still mainly used in research institutions, dance performances and as interactive service guides rather than factory floors or homes.
Tang Wenbin, founder of embodied intelligence startup Yuanli Lingji, recently described the industry as still being at a very early stage. He said the main problem is that robots’ “IQ is too low,” with weak models and low success rates. He added that much of what people currently see is “dancing disguised as working,” suggesting that the public demonstrations often overstate actual commercial readiness.
Still, the Chinese government is heavily backing embodied intelligence, also known as physical AI, as part of its strategy to improve manufacturing productivity and automation. Companies are investing heavily in collecting high-quality real-world training data, often by fitting human workers with sensors and deploying more humanoid robots into factories. For example, UBTech Robotics increased the number of humanoid robots in factories from fewer than 10 in 2024 to more than 1,000 in 2025 and now plans to launch 10,000 full-size humanoid robots for commercial settings this year.
The robot half-marathon may be a spectacle, but it also reflects a larger race China’s push to lead the world in humanoid robotics. While the finish line for true commercial adoption remains far away, events like this show how quickly the technology is moving and how seriously China is investing in the future of intelligent machines.
