Glory Robot Smashes Human Half Marathon Record — But Don't Get Too Excited Yet

April 19th, Beijing Yizhuang. The 2026 Humanoid Robot Half Marathon.

When the Glory Qitian Daxisheng team’s robot “Lightning” crossed the finish line, I was at home watching the stream. My first reaction was to double-check the time. Not because the robot was wrong — because I couldn’t believe what I was seeing.

50 minutes, 26 seconds.

That’s nearly 7 minutes faster than the human half marathon world record.

Honestly, my first thought wasn’t “AI is incredible.” It was: there might be more here than meets the eye.

Let me start with the technical reality.

21 kilometers, 50 minutes 26 seconds, averaging 2 minutes 24 seconds per kilometer. For a humanoid robot, this is genuinely impressive engineering. But there’s a key detail everyone seems to be glossing over: this half marathon was run on a closed, flat, pre-mapped course. No traffic lights, no complex terrain, no crowd interference. The robot ran on visual positioning and pre-loaded maps the entire way — in some sense more like “a machine on a treadmill” than an actual human runner.

This isn’t dismissing the Glory team’s technical achievement — getting a humanoid robot to run 21km at all is an engineering miracle. What I’m saying is: comparing this directly to the human marathon record is media clickbait, not science.

Now, the financial angle.

When Honor — a phone manufacturer — decides to build robots, what’s the first reaction? “They’re straying from their core business, right?”

I think it’s actually the opposite. Phone manufacturers have the most ready-made落地 scenario for edge AI, the ability to design their own chips, and existing user data. If Honor could put a “robot control mode” into their phones, turning the device into a robot’s computing hub — that’s a way more interesting story than selling another smartphone.

The global smartphone market is saturated. Every manufacturer is hunting for the next growth lever. Embodied intelligence might be the most compelling narrative Honor can tell right now.

Finally, the industrial perspective.

Last year’s robot half marathon best time: 2 hours 40 minutes, with only 6 out of 20 teams finishing. This year: over 70 teams, and even the worst finishers significantly outperformed last year’s winner.

That rate of improvement is what’s actually terrifying.

At this pace of evolution, next year’s robot half marathon will likely outpace the human record by 15+ minutes. When that happens, the media headline won’t be “robots run faster than humans.” It’ll be “humans can’t keep up with robots.”

But the question I actually care about: what can these robots do tomorrow that they can’t do today?

If “Lightning’s” technology can transfer to logistics robots or home service robots, that 50 minutes 26 seconds might be worth far more than its headline number suggests.

Let’s not get ahead of ourselves. Let’s see where this goes.

— Lin Rui, writing from Shenzhen

Your turn: Which embodied AI application will hit the mainstream first — industrial, logistics, or home? Drop a comment.