Last Sunday morning in Beijing's E-Town district, a humanoid robot built by Chinese smartphone brand Honor crossed the finish line of a half-marathon in 50 minutes and 26 seconds. That is faster than any human being has ever run the distance. Ugandan runner Jacob Kiplimo set the human world record just last month in Lisbon. Honor's robot beat it without a single day of physical training, without tapering, and without needing to eat anything at the finish line.
The headline number is striking, but the more revealing figure is the year-on-year comparison. In last year's inaugural Beijing Humanoid Half Marathon, the winning robot finished in 2 hours and 40 minutes. That is more than double the human winner's time in the same race. In twelve months, the gap closed from "embarrassingly slow" to "faster than any human who has ever lived." The improvement curve on humanoid locomotion is not gradual. It is a cliff.
This year's race drew over 100 competing teams, up from 20 last year. Nearly half the robots navigated the entire course autonomously, without remote operators. That matters more than the speed record in some ways: a robot being steered by a human around a road course is essentially a very expensive remote-control car. An autonomous robot that manages its own balance, pacing, and route decisions over 21 kilometres is something categorically different. Honor's winning robot was autonomous. A faster Honor machine actually finished in 48 minutes and 19 seconds, but was disqualified from the top prize because it was remote-controlled.
The race was as much a technology showcase as a sporting event. China's robotics sector has seen heavy state backing alongside aggressive private investment, and the field of competitors at E-Town reflected that: the podium was swept entirely by Honor robots, but the broader participation included hardware from dozens of companies working across different mechanical architectures and control approaches. Some robots fell at the starting line. One hit a barrier. These failures are not embarrassments; they are what a genuinely competitive development environment looks like from the outside.
The social media reaction to the result was telling. One widely circulated comment read: "My car can outrun a cheetah too." It is a fair observation in a narrow sense: a machine optimised for a task will outperform biology at that task, and this has been true since the first steam engine. But the comparison misses the point of what makes humanoid robots interesting. Cars are not general-purpose. A robot that can run a half-marathon autonomously is running on legs, processing sensor data in real time, and recovering from perturbations in a dynamic environment. The same architecture, refined, can navigate a warehouse, climb stairs, or assist in a hospital. Speed records are the headline. General mobility is the actual story.
For Beijing, the race served a second function: demonstrating publicly that China's robotics industry has moved past the "proof of concept" phase. The government has been explicit about humanoid robots as a strategic technology priority, and events like E-Town give companies a public, comparable benchmark to test against. The fact that the improvement from last year was this dramatic will not be lost on the industry's investors or its competitors abroad.
The human half-marathon, for the record, was also held at E-Town. China's Zhao Haijie won the men's race in 1:07:47. Wang Qiaoxia took the women's title in 1:18:06. Both times are excellent. Neither will be in the headline.