Video: The Second World Humanoid Games were held in Beijing
Humanoid robots have learned to run the 100-meter dash, play soccer, dance, and perform complex physical tasks. During the Second World Games for Humanoid Robots in Beijing, some robots even achieved results that surpassed human world records.
The competition took place August 22–26, 2026, and brought together more than 2,000 robots from 666 teams.
The robots competed in 51 disciplines, including track and field, soccer, boxing, tennis, weightlifting, and other sports.
One of the main highlights of the competition was the speed of the humanoid robots.
The Tiangong Ultra robot initially ran 100 meters in 9.39 seconds, surpassing Usain Bolt’s world record of 9.58 seconds. Later, the robot improved its own time to 8.86 seconds, and in the final, it clocked 8.64 seconds.
Thus, based on the official result, the machine already significantly outperforms the best human performance. At the same time, the comparison is not entirely fair: robots and humans have different physiology, design, and movement conditions.
Tiangong Ultra has also demonstrated results that surpassed human records in other track and field events.
[video width="640" height="640" mp4="https://cdn-cabinet.ua.news/uploads/images/img-5893.mp4"][/video]
Soccer was a separate part of the competition, where humanoid robots had to not only move across the field but also coordinate their actions, locate the ball, and interact with other machines.
[video width="848" height="476" mp4="https://cdn-cabinet.ua.news/uploads/images/img-5891.mp4"][/video]
This is more important than simply demonstrating speed, as team sports require robots to navigate their surroundings, react quickly, and coordinate their movements.
[video width="464" height="824" mp4="https://cdn-cabinet.ua.news/uploads/images/img-5898.mp4"][/video]
During the games, the machines also competed in tennis matches, boxing, dancing, and other disciplines.
These impressive results do not mean that humanoids are already ready to work alongside humans in challenging conditions.
[video width="848" height="476" mp4="https://cdn-cabinet.ua.news/uploads/images/img-5889-1.mp4"][/video]
During the races, some robots lost their balance after crossing the finish line, crashed into safety mats, or broke down. Falls also occurred during other events at the competition.
That is why it is important for engineers to focus not only on maximum speed but also on the ability to stop, maintain balance, adapt to unpredictable situations, and interact safely with people.
[video width="640" height="640" mp4="https://cdn-cabinet.ua.news/uploads/images/img-5892.mp4"][/video]
The competition in Beijing effectively served as a major testing ground for physical artificial intelligence—systems that must not only analyze information but also act in the physical world.
The organizers included not only sports disciplines in the program but also practical scenarios, such as household chores, firefighting, and assisting in retail.
[video width="464" height="824" mp4="https://cdn-cabinet.ua.news/uploads/images/img-5898-1.mp4"][/video]
For developers, such tasks may be even more important than sports records. A robot that can run 100 meters quickly isn’t necessarily capable of working in a factory, a warehouse, during an emergency, or in other challenging conditions.
It is precisely the transition from a controlled sports arena to a real-world environment that remains one of the main challenges for manufacturers of humanoid robots.
The main goal of such competitions is not to replace the Olympics for humans, but to accelerate the development of robots capable of moving autonomously, making decisions, and performing physical work.
China is actively investing in the production of humanoid robots and aims to transfer technologies from sports arenas to manufacturing, logistics, the service sector, and other industries.
So the current records are just part of the picture. The most important indicator of progress will be how well humanoids learn to operate in environments where there are no ideal surfaces, clear rules, or pre-programmed scenarios.
Currently, Chinese tech companies are accelerating the development of so-called “world models”—artificial intelligence systems designed to help humanoid robots understand and interact with their physical environment. At the World Conference on Artificial Intelligence in Shanghai, hundreds of robots demonstrated boxing, making coffee, sorting objects, and other tasks; however, outside of these prepared demonstrations, their capabilities remained limited.
