Running · Published
A robot ran a half-marathon in 50:26. Read the fine print before the headline.
From 2:40:42 to 50:26 in one year is a large jump. What changed, and what the two times cannot tell you.
Sport
Source-backed robot sport reporting and results: running, football, basketball, combat, tennis, table tennis and badminton. Humanoids and specialist robots are identified separately.
This is a selected evidence file, not a complete current world ranking. A source link is not independent verification.
Desk reporting
Football · Published
A Chinese team wins RoboCup AdultSize for the first time, on a shared robot
Tsinghua and China Agricultural University met in the 2025 final, both on Booster T1 robots.
Basketball · Published
The specialist shooter: what Toyota's CUE records do and do not show
Two Guinness-certified marks, five years apart, for a robot built to do one thing.
Boxing & combat · Published
Why this desk has no humanoid boxing leaderboard yet
Fights have happened since May 2025. A ranking would still imply a measurement nobody has made.
Tennis · Published
A humanoid can return a tennis ball. That is not the same as winning a match.
LATENT's Unitree G1 reports 90.90% forehand and 77.78% backhand return success under externally tracked research conditions.
Table tennis · Published
Thirteen wins, twenty-nine matches: what robot table tennis actually measured
DeepMind's 2024 study offers a match-based result, but its robot is a specialist arm and its opponent mix is essential context.
Badminton · Published
Badminton's useful number is 42 of 46 hardware returns—not a simulated rally
A humanoid research platform coordinates footwork and striking, but motion capture and a restricted hitting area remain important limitations.
Running results
Autonomous and remote-controlled runs are charted separately. Organisers applied scoring adjustments in some events, so a scored time is not always the raw finishing time.
Humanoid half-marathon: winning times by edition
Finishing time, 21.0975 km course · seconds
Historical winners, not a head-to-head ranking. The 2026 winner used autonomous navigation; the 2025 team described wireless-lead following navigation and three battery changes. Rules, hardware and operating conditions differ across editions.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| Honor “Lightning” (Qitian Dasheng team) | 0:50:26 | 2026-04-19 | Autonomous navigation | Source-reported result | Reported by event organisers as the winning time. Outlets used the names “Lightning” and “Flash” for the winning Honor robot. PBS News / Associated Press |
| Tiangong Ultra (Beijing Humanoid Robot Innovation Center) | 2:40:42 | 2025-04-19 | Not specified / not applicable; see notes | Organiser or state-media reported | 2025 winner. The team described wireless-lead following navigation and three battery changes; not assumed equivalent to the 2026 autonomous-navigation category. Beijing municipal government portal |
Humanoid half-marathon: first across the line, remote-controlled
Finishing time, 21.0975 km course · seconds
Single entry. A remote-controlled robot is piloted by people, so it is not ranked against autonomous runs.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| Honor remote-controlled robot | 0:48:19 | 2026-04-19 | Remote-controlled | Organiser or state-media reported | Global Times, relayed by the Associated Press, reported this robot crossed first but did not take the championship, which went to an autonomous entry under weighted scoring rules. PBS News / Associated Press |
World Humanoid Robot Games 2025: 1,500 m winner
Winning time · seconds
2025 winner only; not a current world ranking. Navigation modes and event scoring must be checked before comparing this mark with autonomous-only runs.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| Lingyi Technology team (Unitree) | 6:34.40 | 2025-08-15 | Not specified / not applicable; see notes | Organiser or state-media reported | Beijing News reported 39 teams, 11 heats and an 84.6% completion rate. Retained at team scope: the cited result names the team, not a hardware revision. Beijing News (新京报) |
World Humanoid Robot Games 2025: 400 m winner
Winning time · seconds
Winner only. Different event from the 1,500 m and half-marathon and not charted against them.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| Gaoyi Technology team | 1:28.03 | 2025-08-15 | Not specified / not applicable; see notes | Organiser or state-media reported | Beijing News reported 59 teams, 19 heats and an 84.7% completion rate. The cited result is retained at team scope, without assigning it to a current robot model. Beijing News (新京报) |
Football results
Football results are placings, not measurements, so the chart records who won each final rather than scoring robots on a shared scale.
RoboCup 2025 Humanoid League, AdultSize: final placings
Final placing (1 = champion) · place
Same tournament and category. Teams used the same Booster T1 platform, so the result reflects team software and play as much as the robot.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| Tsinghua Hephaestus (Booster T1) | 1st | 2025-07-21 | Not specified / not applicable; see notes | Official event result | First Chinese team to win the category. Xinhua reports the final was won against China Agricultural University. RoboCup Humanoid League official results |
| CAU Mountain&Sea (Booster T1) | 2nd | 2025-07-21 | Not specified / not applicable; see notes | Official event result | Runner-up in an all-Chinese final, as reported by Xinhua and Beijing Daily. RoboCup Humanoid League official results |
Basketball results
These are certified records for one task, shooting. They are not a measure of how a robot would play a game. Note that “assisted” is a category label in the Guinness title.
Farthest basketball shot by a humanoid robot
Shot distance · metres
Single certified record entry.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| Toyota CUE6 | 24.55 m | 2024-09-26 | Not specified / not applicable; see notes | Record body certified | Achieved in Nagakute, Japan, according to the Guinness World Records entry. Guinness World Records |
Most consecutive free throws by a humanoid robot (assisted)
Consecutive successful free throws · throws
Single certified record entry in the Guinness “assisted” category.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| Toyota CUE3 (with Alvark Tokyo) | 2,020 | 2019-05-17 | Not specified / not applicable; see notes | Record body certified | Achieved in Fuchu, Japan, per the Guinness World Records entry. The record is older than the other charts on this desk. Guinness World Records |
Boxing & combat results
No cross-event ranking yet. The sources reviewed describe human-machine collaboration and mixed formats; they do not establish comparable standings across events or robot platforms.
No comparable cross-event standings published in this file. No ranking is inferred.
Tennis results
The selected evidence covers LATENT's March 2026 research evaluation, not a current world ranking. Forehand and backhand returns are separate measurements, not match wins. External tracking and test conditions matter.
LATENT on Unitree G1: real-world forehand returns
Balls successfully returned inside the opponent's court · %
One research system and one stroke type, not a ranking of robots. The authors report 20 human–robot rally evaluations; the per-stroke sample counts are not separately stated. Evidence date is the paper submission date, not a known test day.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| LATENT · Unitree G1 (forehand) | 90.90% | 2026-03-13 | Autonomous control; see tracking and setup notes | Source-reported result | Research-author reported hardware result, Table 5. External optical motion capture tracks the marked ball and robot; a racket replaces the right hand. This is a return task, not a full tennis match or an onboard-vision result. LATENT paper · Table 5 |
LATENT on Unitree G1: real-world backhand returns
Balls successfully returned inside the opponent's court · %
One research system and one stroke type, not a ranking of robots. The authors report 20 human–robot rally evaluations; the per-stroke sample counts are not separately stated. Evidence date is the paper submission date, not a known test day.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| LATENT · Unitree G1 (backhand) | 77.78% | 2026-03-13 | Autonomous control; see tracking and setup notes | Source-reported result | Research-author reported hardware result, Table 5. External optical motion capture tracks the marked ball and robot; a racket replaces the right hand. This is a return task, not a full tennis match or an onboard-vision result. LATENT paper · Table 5 |
Table tennis results
This selected result is a specialist robotic arm, not a humanoid. Match wins against a particular mix of human opponents cannot be compared directly with stroke-return rates or a different study's players.
Google DeepMind: competitive matches against unseen human players
Matches won in the published mixed-skill evaluation · matches
One 2024 research evaluation of 29 matches against players ranging from beginners to tournament level. This is not a tournament placement or a cross-system league. Evidence date is the publication date.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| Google DeepMind · ABB IRB 1100 arm on linear gantries | 13 / 29 wins (45% reported) | 2024-08-07 | Autonomous control; see tracking and setup notes | Source-reported result | Research-author reported. Won all beginner matches and 55% against intermediate players; lost all against the most advanced players. Specialist arm with camera-based ball tracking and motion capture of the opponent's paddle. Study rules and opponent mix are part of the result. Google DeepMind · research publication |
Badminton results
These are two controllers on the same research platform, not two competing robot products. Hardware returns are kept separate from simulated rallies. Dates identify the cited paper version, not a known test day.
Humanoid badminton: controller comparison on machine-fed shuttles
Valid returns in 46 real-world trials per controller · %
Same 21-DoF humanoid and machine-fed hardware test setup in the September 2026 paper version. Compare these controller variants only; neither number is a human-match win rate.
| Competitor | Result | Evidence date | Control mode | Classification | Notes and source |
|---|---|---|---|---|---|
| 21-DoF research humanoid · target-known controller | 91.3% (42 / 46) | 2026-09-14 | Autonomous control; see tracking and setup notes | Source-reported result | Research-author reported hardware result: 42 valid returns and 4 racket-frame hits. External motion capture plus an Extended Kalman Filter supplies the target. The same policy's 21-hit robot–robot rally was in simulation, not on hardware. Humanoid badminton paper · hardware evaluation |
| 21-DoF research humanoid · prediction-free controller | 71.7% (33 / 46) | 2026-09-14 | Autonomous control; see tracking and setup notes | Source-reported result | Research-author reported hardware result: 33 valid returns, 9 frame hits and 4 misses. Removes the explicit trajectory predictor, not external sensing; motion capture still supplies shuttle positions. Humanoid badminton paper · hardware evaluation |
Continuing coverage from the news archive
Message, Market, Margin: How Humanoid Makers Shape Their Sales Playbooks for Logistics, Clinics and Homes
Introduction Humanoid robotics companies are not fighting a single market battle; they are running multiple, staggered go-to-market plays across logistics floors, industrial lines, clinics and living...
The Story Behind the Stride: How Humanoid Makers Reframe the Same Biped for Warehouses, Wards and Living Rooms
Introduction The humanoid robot market in 2025–26 is less a single race than a set of targeted sprints: the same bipedal hardware is being reframed repeatedly to answer different buyer problems. That...
Updates from April to May 2025: The Humanoid Robot Race Accelerates
The global race to develop and deploy humanoid robots has accelerated dramatically since April 2025. Major players are announcing production targets and new capabilities...