
The second World Humanoid Robot Games concluded on August 26 after 1,301 matches. Final Beijing reporting now makes it possible to separate preliminary performances from final results and to document the event’s autonomy rules, scenario tasks and competition records more precisely.
Introduction
The second World Humanoid Robot Games, officially 第二届世界人形机器人运动会 in Chinese, ran from August 22 to 26, 2026 at Beijing’s National Speed Skating Oval. The final field contained 2,056 robots and 666 teams from 16 countries across six continents.
The competition also exposed current limitations. Beijing’s closing coverage reported that autonomous systems could lose substantial performance when objects or environments changed in real-world scenario tasks. That gap between controlled movement and robust task completion is central to the Physical AI evidence from the event.
Research date September 29, 2026. Information verified from official sources available as of September 29, 2026.
Direct answer
The World Humanoid Robot Games 2026 ran August 22 to 26 at Beijing’s National Speed Skating Oval with 2,056 robots, 666 teams from 16 countries, 51 events and 1,301 matches. Tiangong Ultra won the large-group 100 m in 8.64 seconds. The large-group 400 m record reached 38.15 seconds and Tianzhuo won the 1500 m in 2 minutes 21.64 seconds.
World Humanoid Robot Games 2026 final key figures
| Item | Final 2026 figure |
|---|---|
| Official Chinese name | 第二届世界人形机器人运动会 |
| Dates | August 22 to 26, 2026 |
| Venue | National Speed Skating Oval, Ice Ribbon, Beijing |
| Robots | 2,056 |
| Teams | 666 |
| Countries | 16 across six continents |
| Total events | 51 |
| Competitive events | 30 |
| Scenario events | 21 |
| Matches | 1,301 |
| Competition units | 9 across five days |
| China | 641 teams and 1,975 robots |
| Chinese participants | 157 companies and 200 universities or research institutes |
| Referees | 223 including 28 international, 75 national and 120 first-class referees |
| Hosts | Beijing Municipal People’s Government, China Media Group, World Robot Cooperation Organization and RCAP International Council |
Selected final 2026 results
| Event | Robot or team | Reported result |
|---|---|---|
| 100 m large group | Tianjiao team, Tiangong Ultra | 8.64 s |
| 400 m large group | Tiangong Ultra | 38.15 s |
| 400 m small group | Tiangong Omni | 45.66 s |
| 1500 m | Tianzhuo team, Tiangong Ultra | 2 min 21.64 s |
| Standing high jump | Tiangong Ultra | 3.40 m competition record |
| 400 m obstacle | Tiangong Omni | 4 min 11.44 s in E-Town reporting; 4 min 11 s in Beijing Daily |
| 5v5 football large group | Tsinghua Hephaestus | Champion |
| Office comprehensive service | Qianxun Intelligence | Champion with 190 points |
Selected results reported by Beijing authorities, not independent TechniaHQ measurements. The jump figure is retained as a reported competition result; the underlying measurement protocol is not established here.
World Humanoid Robot Games 2025 vs 2026
The 2025 figures are final post-event totals. The 2026 figures and results use post-event Beijing reporting through September 3, 2026.
| 2025 | 2026 |
|---|---|
| 280 teams | 666 teams |
| More than 500 robots | 2,056 robots |
| 16 countries across five continents | 16 countries across six continents |
| 26 events | 51 events |
| 487 matches | 1,301 matches |
| 3 competition days | 5 competition days |
| 100 m official scored result 21.50 s | 100 m large-group final 8.64 s |
| 400 m 1 min 28.03 s | 400 m small group 45.66 s and large group 38.15 s |
| 1500 m 6 min 34.40 s | 1500 m 2 min 21.64 s |
| Mixed control modes with a 0.8 scoring coefficient for selected autonomous track entries | Most competitive events required full autonomy. Scenario tasks used separate autonomy score weights |
The 2025 100 m result includes an autonomy coefficient of 0.8. Comparing 21.50 s directly with 8.64 s does not measure a like-for-like speed improvement. The obstacle races were exceptions to the 2026 autonomy rule.
Venue and scenario settings
The Games were held at the National Speed Skating Oval in Beijing. The venue is also known as the Ice Ribbon and hosted the inaugural Games in 2025.
Beijing’s official event listing gives the venue address as 2 Lincui Road in Chaoyang District. The 2026 program also extended scenario competition into nine real-world settings around Chaoyang rather than keeping every task inside a simulated hall.
Final athletics results replaced preliminary times
Several figures published during the competition were preliminary. Tianzhuo’s 9.39-second 100 m run came during the opening program. The large-group final was later won by Tianjiao team’s Tiangong Ultra in 8.64 seconds.
Tiangong Ultra also produced the 38.15-second large-group 400 m result. Tiangong Omni won the small group in 45.66 seconds. Tianzhuo won the 1500 m in 2 minutes 21.64 seconds. Beijing’s closing report lists these as major improvements over the 2025 competition records.
The 2026 rules raised the autonomy requirement
The August 13 official rules state that every track race and other competitive event had to be completed fully autonomously except the 100 m obstacle and 400 m obstacle races.
Scenario events used a separate scoring rule. A task completed fully autonomously received a final-score weight of 1 while a remotely controlled completion received 0.5.
These rules still do not support a claim of general autonomy. A robot can satisfy one competition rule with a task-specific controller and still require human intervention or fail when the environment changes.
21 scenario events tested work in real settings
In the office comprehensive service final, robots had to arrange meeting materials, load and operate a printer, deliver printed documents and feed documents into a shredder. Qianxun Intelligence won with 190 points. Shenzhen Zhiyuan and Ningxia Zhiyuan finished second and third.
Dexterous-hand tasks and control modes
Eight dexterous-hand tasks included tool assembly, powder weighing, block building, nail fixing, bottle opening, unpacking, picking beans with tweezers and cable connection.
Beijing reporting on the finals describes teams adjusting perception for lighting changes and choosing between full autonomy and remote control. These tasks expose errors in visual localization, fingertip positioning, force control and recovery that are easy to hide in large-object handling.
The hand finals report identifies four autonomous winners among the eight events, while also describing remote operation. That is why the track-autonomy rule cannot be applied to every hand result.
The training center produced useful pre-event data
By August 13, the humanoid robot training and evaluation base at the venue had operated for 30 days and hosted nearly 1,000 training sessions. The organizer’s robot management platform had connected all 666 teams and 2,056 robots.
The system used dedicated encrypted communication links and tracked operational data such as joint temperature, motion state and battery level. The training period remains useful context because teams exposed locomotion, manipulation and scenario-task problems during repeated practice before formal competition began.
Technology behind the competition
Locomotion
Beijing E-Town reporting says Tiangong Ultra was upgraded in joint motors, mechanical configuration, thermal management, lightweight design and aerodynamics. Training also differed by event. The 100 m emphasized acceleration while the 400 m and 1500 m emphasized cornering speed and sustained control.
Balance and recovery
Fast cornering and obstacle contact create disturbances that a preplanned trajectory cannot ignore. The Tiangong team reported work on high-speed turning control because loss of control at more than 10 m/s could throw the robot outward on the bend.
A dataset announcement is not a checked release
Beijing's closing report announced a dataset covering more than 2,500 hours from training and competition, with 12 application settings and 44 operations. It says access will be provided through the organizing committee. This review has not verified a download, license, sensor schema or usable trial-level release; the announcement alone does not supply those details.
Limitations and missing information
- This page includes selected results supported by official event or Beijing government sources. It does not claim to reproduce the complete medal table for all 51 events.
- The final field contained 2,056 robots but the sources do not describe them as 2,056 identical general-purpose humanoids. Hardware, size classes and task specialization varied.
- Full autonomy in a competition event is a rule-defined property for that event. It does not establish general autonomy across unrelated tasks.
- The 100 m obstacle and 400 m obstacle races were exceptions to the broad 2026 competitive-event autonomy requirement. Scenario tasks could also be completed by remote control with a lower score weight.
- The public official source set confirms 16 countries but does not provide a complete stable list of all 16. Unverified country names were not added.
- No complete official participant roster was available in a stable source during this update. Prominent companies are not labeled as participants unless an official source links them to the Games.
Sources and methodology
The comparison separates the 2025 adjusted 100 m result from 2026 results and distinguishes track rules from scenario scoring. Results and team engineering statements are attributed to the linked reports. No competition footage or dataset was independently evaluated in this review.
Related TechniaHQRobot guides
Fact-check report
Verified
Share this article
Share the current TechniaHQRobot article page.
Continue reading
Open the latest robotics reporting, Physical AI analysis and hardware notes.