Tiangong Ultra Humanoid Outruns Usain Bolt at Robot Games

The second World Humanoid Robot Games in Beijing, which concluded on August 26, gathered more than 600 teams and over 2,000 robots from 16 countries. The event highlighted remarkable athletic benchmarks. The Tiangong Ultra, developed by the Beijing Humanoid Robot Innovation Center, completed the 100-meter sprint in 8.86 seconds and achieved a 7.97-meter long jump. Another humanoid named Lightning, developed by smartphone manufacturer Honor, clocked a 9.47-second sprint at the games, following a 9.32-second run in pre-event testing. Both machines surpassed Usain Bolt's human world record of 9.58 seconds.

These physical feats rely on compact, high-power motors that maintain dynamic balance at high speeds, alongside control algorithms trained in simulation. Roboticist Stefanie Tellex of Brown University noted that China has achieved "amazing success building robot hardware," pointing to Hangzhou-based Unitree as a key innovator. Because these algorithms optimize for stability in virtual environments rather than human biology, the sprinting humanoids often exhibit highly nonhuman running gaits.

For robotics practitioners, however, the real frontier lies in manual dexterity rather than raw speed. The games introduced new challenges like plugging cables into sockets, block building, and tweezer bean picking to test fine motor skills. These tasks illustrate Moravec's paradox, where complex physical maneuvers like backflips prove easier for AI to master than basic object manipulation. Frictional forces and slippage make these precise tasks highly unpredictable for robotic hands.

While many demonstrations at the games relied on human teleoperation, some companies showcased autonomous capabilities. Galbot entered the Supermarket Scenario Competition to test shelf-stocking and demonstrated an autonomous robot playing tennis against a human. For developers, transitioning from simulated environments to unpredictable, real-world applications remains the ultimate goal. These games prove that while hardware is rapidly maturing, the software governing autonomous physical interaction is the next critical bottleneck to solve.