If you thought AI was all about intelligence and not about muscle, think again. AI-powered robots just ran half marathon in China and the results show that the machines are not poised to defeat humans on track and field as well. Running in Beijing in a 21-kilometre race, many humanoid robots not just finished the race but some of them also did so under one-hour mark, around which humans have been stuck forever.
The race saw more than 100 humanoid robots take on the 21.1-kilometre course alongside human participants. While both groups ran the same route, organisers kept them separated by barriers, creating a controlled comparison of speed, balance and stamina.
And crossing the finish line ahead of everyone was a humanoid robot, beating its human competitors by a clear margin.
The standout performance came from a robot developed by Chinese smartphone maker Honor, which finished the race in 50 minutes and 26 seconds. That put it well ahead of the fastest human runner, who crossed the line in 1 hour, 7 minutes and 47 seconds, according to Reuters. On paper, the robot’s timing also surpasses the current human half-marathon world record, although experts caution that such comparisons are not strictly like-for-like.
The win of the robot is a big comeback for machines from last year’s race, where several robots struggled to finish, losing balance or requiring assistance. This time, many machines maintained their pace and completed the course with far fewer issues. According to the report, the event highlighted how quickly the technology is evolving.
A significant number of robots demonstrated improved balance, coordination and endurance compared to earlier attempts.
Nearly 40 per cent of the robots in the race reportedly ran fully autonomously, relying on onboard sensors and AI systems to navigate the course, maintain balance and adjust their stride in real time. Robots that were remotely controlled faced time penalties, while additional penalties applied for battery swaps or full replacements during the race.
Although, machines are still far from perfect. Even the top-performing robot struggled with the limits of the technology. It briefly clipped a railing just metres before the finish line, a small but telling reminder that precision and consistency still need refinement.
According to engineers much of the progress of these humanoids comes from designing robots that mimic human movement more closely. Du Xiaodi, an engineer with the Honor team, told Reuters that the robot’s leg structure — measuring between 90 and 95 centimetres — is optimised for stride efficiency, similar to elite runners. The system also uses liquid cooling technology to prevent overheating during prolonged exertion.
Spectators at the human-versus-machine race noted a visible improvement in how the robots moved. Many appeared more stable and fluid, with running styles that more closely resembled human motion. For some, the race also offered a glimpse into future career paths. Chu Tianqi, a 23-year-old engineering student, told Reuters that the performance of robots reflected the rapid pace of AI development, while younger attendees expressed interest in studying robotics.
Experts say the results should be viewed in context even if the robots were impressive. Running a race — even a long one — is still a controlled task compared to real-world environments. They note that for now humanoid robots are still facing challenges in areas such as fine motor skills, adaptability and decision-making in unpredictable conditions.
