Contents
- The three numbers, answered straight
- 9.39 seconds: Tiangong Ultra's race time
- 9.47 vs 9.32: two numbers, one robot
- Why these beat Bolt's 9.58 — and why that's not a world record
- The crash clip everyone's asking about
- Inside the World Humanoid Robot Games, by the numbers
- FAQ
- What's still unknown
One of the three numbers everyone’s citing from this week’s Beijing robot race isn’t a race time at all. 9.39 and 9.47 are official results from Saturday’s 100m; 9.32 is a pre-event test run that most headlines lump in with the other two anyway. Here’s which number means what, and why “robot beats Bolt” is doing more work than the fine print supports.
The three numbers, answered straight
Tiangong Ultra ran the 100m in 9.39 seconds. That’s the headline figure, reported by Engadget, the South China Morning Post, and Al Jazeera as the result from the actual race.
Lightning, a humanoid built by phone maker Honor, ran 9.47 seconds in that race. ESPN’s coverage refers to the same robot as “Blitz” — more on that naming mismatch below.
A third number, 9.32 seconds, also belongs to Lightning. It’s not the race time. It’s from a preparatory test run, reported separately with a peak speed figure of 14.5 meters per second. Mixing it into “the race” is the single most common error in how this story spread.
| Robot | Time | Type | What it means |
|---|---|---|---|
| Tiangong Ultra | 9.39s | Official race | Main 100m result, built by the Beijing Humanoid Robot Innovation Center |
| Lightning (Honor) | 9.47s | Official race | Same race, same day, second place |
| Lightning (Honor) | 9.32s | Unofficial test | Pre-event trial run, not the race clock |
| Unitree robot | 12.41s | Official race | Third entrant in the same field |
9.39 seconds: Tiangong Ultra’s race time
Tiangong Ultra’s 9.39-second run is the number every outlet is leading with. It comes from the robot built by the Beijing Humanoid Robot Innovation Center — ESPN’s report also uses the brand name “X-Humanoid” for what appears to be the same entity, though that overlap isn’t spelled out clearly in any single source.
Last year’s event, the first edition of the games in 2025, saw Tiangong Ultra run the same 100m in 21.50 seconds. Cutting that down to 9.39 seconds in twelve months is a large jump. No source explains the engineering behind that improvement in detail.
Tom’s Hardware’s headline attaches a top speed to this run: 23.8 mph. That figure lines up almost exactly with the average speed you’d get from dividing 100 meters by 9.39 seconds — about 10.65 meters per second, or 38.3 km/h. That match is close enough that it’s likely an average-speed conversion, not a separately measured peak — but Tom’s Hardware’s full article sits behind a signup wall, so it can’t be confirmed from the original text. Treat 23.8 mph as probably average, not confirmed peak.
9.47 vs 9.32: two numbers, one robot
Lightning is the Honor-built robot in this story, standing 169cm tall. During a sprint, its legs extend from a resting 95cm to 105cm — a stride-lengthening trick rather than a change in stated top speed.
That’s where the two Lightning numbers split apart. The 9.47-second time is from the actual 100m race that Tiangong Ultra won. The 9.32-second time is not. It comes from an earlier preparatory test, paired with a peak speed of 14.5 m/s — roughly 52.2 km/h, or about 32.4 mph.
So Lightning was, on paper, faster in a test than it was on race day. That’s not unusual for any athlete or machine. It is unusual that headlines strung 9.32 next to 9.39 and 9.47 as if all three came from the same event.
Lightning has one other credential worth noting: a half marathon, 21km, completed in 50 minutes 26 seconds. A separate crash was reported during that event too, though details on it are thin.
Why these beat Bolt’s 9.58 — and why that’s not a world record
Usain Bolt’s 100m world record is 9.58 seconds, set in 2009. Sources report the year but don’t consistently name the meet in their text.
All three robot numbers — 9.39, 9.47, and even the unofficial 9.32 — are faster than 9.58 seconds. That gap is the entire basis for headlines like Engadget’s “humanoid robots have beaten Usain Bolt’s 100-meter dash record” — the same framing behind earlier coverage of the Beijing robot games’ 100m record claim.
Here’s the part almost nothing in the coverage states outright: this is not a certified athletics record. Bolt’s 9.58 is a World Athletics-ratified mark, timed and verified under IAAF rules. The robot times come from a robotics exhibition’s internal timing, not from any athletics governing body. It’s worth stating plainly: a faster stopwatch reading is not the same thing as breaking a ratified record.
One more asterisk: it’s unconfirmed whether Tiangong Ultra’s run was a qualifying heat or a final. Search summaries reference something like a “preliminary heat,” but the original wording hasn’t been directly verified.
The crash clip everyone’s asking about
There’s video. A robot goes down.
Engadget, SCMP and Tom’s Hardware all describe Tiangong Ultra colliding with padding — called a “padded wall” in one account, a “safety mat” in another — right after finishing the 100m, then skidding off the course. Engadget also reported robots being carried off on stretchers afterward.
Separately, Interesting Engineering reported a different, more dramatic incident: around August 20–21, 2026, during training, a humanoid robot whose model wasn’t named ran full-speed into a cushioned wall, bent at the waist to nearly a right angle, and threw off sparks at the point of impact. That clip drew more than 13 million views and sparked online arguments about whether the robot’s vision or braking system failed.
Are these the same incident filmed from different angles, or two separate crashes? Unconfirmed. The sourcing doesn’t resolve it either way.
Inside the World Humanoid Robot Games, by the numbers
This is the second edition of the World Humanoid Robot Games, held in Beijing. The first ran in August 2025.
The 2026 race took place on Saturday, August 22, though one source’s wording muddles it with the 22nd–23rd. The full event ran five days.
Scale, per the sourcing:
- More than 2,000 humanoid robots entered, per ESPN
- Interesting Engineering breaks that down further: about 2,056 robots
- 666 teams total
- 157 Chinese companies represented
- 641 Chinese teams
- 1,975 Chinese-built robots
- 16 countries participating
- 51 sporting events on the program
- More than 1,000 individual sub-events
- Roughly 3,000 products on exhibition
- Robots marched in formation by the hundreds at the opening ceremony
ESPN alone names the venue as the Beijing National Speed Skating Oval — a claim not cross-checked against any other outlet in this reporting, so treat it as single-sourced.
The 100m wasn’t the only event where a robot outran a human benchmark. In the high jump, an X-Humanoid robot cleared 2.88 meters. The human world record there is 2.45 meters, held by Cuba’s Javier Sotomayor since 1993. Same caveat applies: an exhibition result, not a ratified athletics record.
A third robot, built by Unitree Robotics, ran the 100m field in 12.41 seconds — slower than both Tiangong Ultra and Lightning, and rarely mentioned in headlines built around the top two. Unitree’s specific model wasn’t named in the sourcing, only “a Unitree Robotics robot.”
FAQ
Did a robot actually break Usain Bolt’s 100m world record? No, not in the sense that phrase usually means. A robot ran the distance faster than Bolt’s 9.58-second time, at a robotics exhibition. Bolt’s mark is a certified athletics record; the robot times are not measured or ratified under the same rules. Nothing in the sourcing directly states this distinction, but it follows from what kind of event this was.
Which number is the “real” one — 9.39, 9.47, or 9.32? 9.39 (Tiangong Ultra) and 9.47 (Lightning) are both official results from the same 100m race. 9.32 is Lightning’s unofficial time from an earlier test run, not the race itself.
Is the Honor robot called Lightning or Blitz? Most outlets — Engadget, SCMP, Al Jazeera — use “Lightning.” ESPN uses “Blitz.” Both likely refer to the same Honor-built robot, but no source explains the discrepancy, so it’s not fully confirmed.
What’s still unknown
- Whether the 23.8 mph figure is a measured top speed or just the average speed worked out from 100m over 9.39 seconds
- Whether the padded-wall crash after the race and the training-ground crash reported by Interesting Engineering are the same event or two different ones
- The correct English name for Honor’s robot — Lightning or Blitz
- Tiangong Ultra’s actual peak speed, which no source states (only Lightning’s 14.5 m/s test figure is on record)
- Whether the Beijing National Speed Skating Oval is really the venue, since only ESPN says so
- Whether Tiangong Ultra’s 9.39-second run was a heat or the final
This is a one-sentence paragraph, and it’s the honest way to end: more of this story is unconfirmed than the headlines let on.
Sources: Engadget, South China Morning Post, ESPN, Al Jazeera, Interesting Engineering, Tom’s Hardware (headline and partial text only).