Galbot robots complete 100-plus autonomous tennis rallies in Beijing
Key Takeaways
- •Galbot's robots sustained over 100 consecutive autonomous tennis rallies during the opening ceremony of the Second World Humanoid Robot Games in Beijing, which the company called the first live autonomous humanoid tennis match.
- •The demonstration used Galbot's G1 humanoid, a 173-centimeter, 85-kilogram machine with a 240-centimeter extended reach, paired with the LATENT software framework trained on roughly five hours of recreational players' motion data through reinforcement learning and large-scale simulation.
- •The second edition of the World Humanoid Robot Games hosted 2,056 machines entered by 666 teams from 16 countries.
- •Tiangong Ultra, built by the government-backed Beijing Innovation Center of Humanoid Robotics, ran 100 meters in 9.39 seconds, beating Usain Bolt's 9.58-second record, and later completed a 400-meter dash in an adjusted 38.15 seconds.
- •Galbot announced the rally record in an August 23, 2026 post on X, while Unitree's claim that its Superman humanoid reached 12.66 meters per second came from a company video and was not independently verified.
Galbot’s human-like robots completed more than 100 consecutive tennis rallies without human control.
The demonstration took place as the Second World Humanoid Robot Games (WHRG) opened in Beijing, and the company described it as the first live autonomous humanoid tennis match.
Five hours of amateur footage trained the G1 to swing
Beijing Galaxy General Robot Co. gave the performance a loaded label, calling it tennis’s “AstraTennis moment,” a reference to AlphaGo’s 2016 victory over Go champion Lee Sedol.
The parallel carries extra weight in China: AlphaGo’s win is widely described as a catalyst for Beijing’s national AI push, which produced the New Generation Artificial Intelligence Development Plan in 2017.
While AlphaGo solved a board game, this contest unfolded on a real court in real time. Galbot said the rally streak set a record for humanoid robot tennis, and that the demonstration was shown on a live broadcast.
The match was held during the WHRG opening ceremony. The second edition of the event, which follows the inaugural World Humanoid Robot Games held in Beijing in 2025, hosted 2,056 machines from 666 teams across 16 countries.
During play, each robot had to process incoming shots and decide whether to use a forehand, backhand, serve, return, or a scrambling recovery shot.
Galbot’s robots also played doubles with human tennis champions. On court, the robots adjusted their positioning and strategies.
When they were knocked off balance during fast exchanges, the machines recovered, regained their posture, and stayed in the point.
Sports showcases have become a preferred proving ground for China’s humanoid sector since the Ministry of Industry and Information Technology issued 2023 guidance naming the machines a priority frontier technology. Athletic events test real-time perception, balance, and whole-body control — the same capabilities that developers such as Unitree Robotics and Tesla’s Optimus program are pursuing for work outside structured factory settings.
Galbot’s G1 is the physical platform used for the demonstration. It is a 173-centimeter, 85-kilogram humanoid with a 190-centimeter arm span and a torso lift that extends its reach to 240 centimeters. Galbot said those dimensions were selected to mimic the way a human would cover a court.
Each arm has 7 degrees of freedom and interchangeable end-effectors, including two-finger grippers and 22-degree-of-freedom dexterous hands. Visual and tactile sensors feed the perception system, which tracks the ball’s speed, spin, and path.
The software, LATENT, stands for Learning Athletic Humanoid Tennis Skills from Imperfect Human Motion Data.
According to Galbot, the framework breaks tennis into reusable motion primitives, including forehands, backhands, lateral shuffles, and crossover steps, then reassembles them into full strokes. The company said the system is designed to learn from coarse amateur data.
Learning from messy human demonstrations rather than hand-scripted movements is a broader shift in robotics, with research teams increasingly training general-purpose humanoids on teleoperation and motion-capture data.
Researchers collected about five hours of tennis motion from recreational players using a small rig. They then used reinforcement learning and large-scale simulation to train the robot to select and combine the right motions.
Tiangong Ultra ran 9.39 seconds, Bolt ran 9.58
Cryptopolitan reported that a headless humanoid named Tiangong Ultra, built by the government-backed Beijing Innovation Center of Humanoid Robotics, ran 100 meters in 9.39 seconds. That time beat the 9.58 seconds Usain Bolt set in 2009.
The same model, which won the world’s first humanoid half-marathon in Beijing in April 2025, later completed a 400-meter dash in an adjusted time of 38.15 seconds.
Cryptopolitan also reported that some runners failed to start or moved at a walking pace, while one boxing robot fell into the ropes and remained still until people carried it away.
At the Beijing venue, one retiree said the races were “just a competition of speed. It’s not a test of intelligence.”
Unitree Robotics said its Superman humanoid reached 12.66 meters per second, although Cryptopolitan noted that the figures came from a company video and were not independently verified.
On August 23, 2026, Galbot posted on X: “WORLD RECORD! Galbot robot completed 100+ consecutive autonomous tennis rallies with zero mistakes! 🎾 Fully autonomous. 📷 Live Broadcast. A new milestone for humanoid robot tennis! #AstraTennis #EmbodiedAI #HumanoidRobots #WHRG #Galbot pic.twitter.com/c7uTrX09ES”
The post was shared from @GalbotRobotics and included the same message about the autonomous tennis rallies.
If you're reading this, you’re already ahead. Stay there with our newsletter .
