The World Humanoid Robot Games is an international competition designed to test humanoid robots across sports, dexterous manipulation and real-world scenario-based tasks. This year’s events ranged from athletics, martial arts and dance to hotel services, library operations, emergency response and industrial applications.

BEIJING, August 26, 2026 — AGIBOT made its debut at the 2nd World Humanoid Robot Games, securing 18 gold, 16 silver and 12 bronze medals. With 46 medals in total, AGIBOT ranked first on both the gold medal table and the overall medal table.

The World Humanoid Robot Games is an international competition designed to test humanoid robots across sports, dexterous manipulation and real-world scenario-based tasks. This year's events ranged from athletics, martial arts and dance to hotel services, library operations, emergency response and industrial applications.

AGIBOT's medal wins spanned dexterous manipulation, scenario-based tasks and motion-control events, including titles in martial arts, dance and obstacle racing. Notably, AGIBOT's competition lineup, including OmniHand, G2, A3 and X2, consists of robots already in mass production or deployed in real-world applications, rather than models developed specifically for competition.
In the dexterous-hand competition, OmniHand reached the finals in all eight events and won seven gold medals, including in Block Building, Powder Weighing and Bean Picking with Tweezers.

OmniHand performs a precision manipulation task
In motion-control events, AGIBOT X2 won gold in the 100m obstacle race, while also securing silver and bronze medals in the 400m obstacle race. AGIBOT A3 also took gold in the Tai Chi event. AGIBOT X2 competed in its mass-produced configuration without hardware modifications specifically for the Games.

AGIBOT X2 competes in the obstacle race
AGIBOT X2 is powered by AGIBOT's self-developed AGILE (AgiBot Generative Intelligent Locomotion Engine) framework, which combines environmental perception with autonomous or semi-autonomous locomotion decision-making across irregular obstacles and varied terrain. Together with precise parameter identification and simulation-to-reality (Sim2Real) transfer, the system is designed to improve stability and robustness when locomotion policies are deployed on physical robots.