Robot 100m in 8.64 seconds, faster than Bolt, but foam padding waits behind the finish line

8.64 seconds, but there's foam padding behind the finish line

On August 22, the second World Humanoid Robot Games opened at Beijing's National Speed Skating Oval, known as the "Ice Ribbon." Tiangong Ultra, built by the Beijing Humanoid Robot Innovation Center (X-Humanoid), ran 9.39 seconds in the first-day preliminary heat for the large-frame 100m category, then 8.86 seconds in the semifinal. It went on to win the final on August 26 with a time of 8.64 seconds. For comparison, Usain Bolt's human world record, set in 2009, stands at 9.58 seconds.

The preliminary, semifinal, and final times belong to different rounds. The 8.64-second mark is the official final result, but equipment, rules, and braking methods differ so much from human competition that it can't be read as a direct comparison of physical capability.

And there's one thing at this race that doesn't exist in human track events: a thick foam pad standing just past the finish line. The robots can't slow themselves down on their own, so they stop by running into the pad. After the 100m preliminaries wrapped up, all three robots that had competed were carried off the track on stretchers.

From 21.50 seconds to 8.64 in a single year

The competition first launched in August 2025. Back then, the winning 100m time was 21.50 seconds, and the winning robot also came from the Tiangong line. The same event, run by the same robot lineage, saw its time cut by more than half in just one year.

Event	2025 (1st Games)	2026 (2nd Games, final)	Human world record
100m	21.50s (Tiangong, winner)	8.64s (Tiangong Ultra)	9.58s

Robot records move this fast because their upgrade cycle doesn't follow the human one. Human records improve generation by generation, but for robots, swapping out joint motor output, batteries, and control algorithms can show up in results the very next season. According to specs published by IT之家, Tiangong Ultra stands 180cm tall and weighs 52kg, and this round's improvements centered on joint torque, rotational speed, battery capacity, and a gait refined through reinforcement learning.

At 8.64 seconds, the robot averaged 41.7 km/h. Alternating footsteps at that speed while holding balance and staying in lane draws on the same underlying capability that delivery robots or factory robots need to navigate around people. But a human 100m race includes both self-powered acceleration and self-powered stopping as part of the event — here, the foam pad handled the deceleration. That's exactly why it's hard to put this number next to Bolt's 9.58 seconds.

The real rule change this year was in the scoring

Last year's 400m winning time was 1 minute 28.03 seconds, set by the Gaoyi Keji team using a Unitree H1. But that result came from remote piloting. At the time, Unitree's founder explained that although the robot had autonomous capability, the team chose remote control for the sake of speed.

This year's Games put scoring weight right on that issue. Gong Xiao, deputy director at the China Software Testing Center, explained that "fully autonomous operation carries a scoring weight of 1.0, while remote-piloted operation carries 0.5." The reasoning behind doubling the gap, she said, is that organizers want robots to become genuinely autonomous partners. In effect, results driven by a human holding a controller and results driven by a robot's own decision-making no longer sit on the same scoreboard.

For reference, this year's large-frame 400m final went to Tiangong Ultra with a time of 38.15 seconds, marking the Games' first gold medal. That's faster than the human world record of 43.03 seconds, but since the manufacturer and control method differ from last year's 400m result, it can't be strung together as "50 seconds shaved off in a year."

Of 51 events, 21 are real-world tasks, not track races

Start with the scale of the event.

Category	2025 (1st Games)	2026 (2nd Games)
Teams	280	666 (+138%)
Robots	~500	2,056 (4x)
Countries	16 (5 continents)	16 (6 continents)
Events	26	51 (30 competitive · 21 scenario)
Total matches	487	1,301
Duration/match units	3 days · 6 categories	5 days · 9 categories
Judges	123	223

What grew was the number of teams and robots, not the number of countries. Sixteen countries participated, same as last year, and of the 666 teams, 641 were from China — spanning 157 companies across 30 provinces and cities. Despite the "World" in its name, the actual makeup looks closer to a domestic Chinese competition, and that's worth keeping in mind when reading these numbers.

The event lineup, though, has clearly changed. Of the 51 events, 21 are scenario-based. Robots are scored on handling actual tasks across nine categories: home cleaning, hotel service, industrial production, emergency response, library management, retail, office service, landscape patrol, and vehicle charging. More than 40% of the Games now judges whether a task got done, not how fast a robot ran.

Zhang Feng, Party Committee Secretary of the Council at the China Electronics Society, said at the opening ceremony that "humanoid robots are evolving from moving on stages and running on tracks to being used in homes and working in factories." According to the "2026 Humanoid Robot Industry Development Report" he presented at the World Robot Conference on August 20, China's humanoid robot shipments topped 40,000 units in the first half of this year.

The track winner and the mass-production leader are different companies

In the large-frame 100m preliminary, a Unitree robot finished third in its heat with a time of 12.41 seconds. Unitree is the company holding the largest volume in this market, having announced on August 12 that its cumulative production of bipedal humanoid robots had surpassed 18,000 units.

Unitree said it scaled back the number of events it entered in order to focus its resources on developing and delivering mass-production models. Tiangong, by contrast, is a platform built by an innovation center established by the Beijing municipal government, and posting strong results at competitions like this one is itself part of its development mandate. Track records and production volume answer two different questions, so reading this race's rankings as a market ranking would be a mistake.

The robot that ran with its hands over its face

One more scene drew attention. Tiangong Omni, from the same innovation center, ran the small-frame 400m with both arms raised in front of its face, swinging its hips side to side as it moved. In China, the move earned the nickname "face-covering run." Tiangong Omni finished the event's final in 45.66 seconds.

Han Gang, a researcher on motion control algorithms at the innovation center, explained what happened. The team originally designed a human-like gait with swinging arms, but during reinforcement learning, the robot shifted into this posture on its own. Swinging the arms quickly concentrates load on the shoulder joints and raises the risk of overheating; keeping the arms near the face relieves that stress and lets the system push more output to the legs instead. In simulation, the team only steered broad targets like speed and heat management — the specific posture wasn't something the developers designed or expected. A rumor that the posture was "learned by watching a girl's move" has been officially denied by the team.

It's a clear example of how a design meant to mimic humans can drift toward whatever works best for the machine once learning takes over. It also lines up with a broader trend: training robots by generating massive amounts of motion data inside simulations is becoming the industry standard.

Editor's take

The number everyone will quote is 8.64 seconds, but the number that will actually move the industry is 1.0 versus 0.5. Autonomous gets full weight, remote-piloted gets half. Set that next to the fact that last year's 400m winner chose remote control for the sake of speed, and it's obvious what the organizers were trying to fix. Every humanoid demo video has drawn the same skeptical question — "isn't someone just controlling that by hand?" — and this year's scoring system answers it directly.

Track records will keep falling. Joint motors and batteries improve every year, after all. The bigger news will be the day organizers no longer need to put foam padding behind the finish line. What homes and factories actually need isn't top speed — it's the ability to stop and get out of the way.

One more thing worth watching: this event is co-hosted by the Beijing municipal government and China Media Group (CMG), among others, and doubles as a showcase for a sector China has designated as a strategic industry. That means records and scale get plenty of airtime, while failures tend to leave a much thinner trail. This competition will become genuinely useful as an industry benchmark once organizers start publishing detailed scoring for the 21 scenario events and reporting how many robots failed to complete their tasks.