A humanoid robot fights a man, Unitree's new $6,500 hand, Figure does some homework, Faraday launches 9 new robots, and more ...
Today in Humanoid Daily ... Unitree launches a new (and good!) hand, Boston Dynamics opens an Atlas factory training ground, Figure's humanoid enters 30 homes it had never seen. Faraday Future launches 9 new robots, a human fights a humanoid robot in a cage and loses ... and the number of the day is 7,000.
Unitree's new robot hand has 22 degrees of freedom and costs $6,500
Unitree just introduced the Dex5-S, a new precision dexterous hand with 22 degrees of freedom. It's not the most of any hand (hello Neo from 1X) but it essentially matches the complexity and proportions of a human hand.
It's 1-to-1 real-hand size, all 22 joints are smoothly backdrivable, and every joint has limit impact torque protection. Translation: the fingers can move naturally, give way when external force is applied, and protect themselves when they hit their mechanical limits.
Price starts at $6,500, before taxes and shipping.
Dexterous hands are becoming one of the most important — and difficult — pieces of the humanoid puzzle. Dancing gets the viral videos, but hands are what actually make a humanoid economically useful. Picking up oddly shaped objects, handling tools, plugging in cables, manipulating small components ... that's where you need far more than a simple gripper.
And 22 degrees of freedom puts Unitree achingly close to human-hand territory.
Boston Dynamics opens an Atlas training lab inside Hyundai's factory
Boston Dynamics announced today the opening of its Robotics Metaplant Application Center at Hyundai's Metaplant America facility in Georgia. The idea: a dedicated space where Atlas humanoids get trained for real manufacturing tasks in a real production environment.
This isn't a demo lab. It's inside an active auto plant. Atlas learns factory tasks the way you'd want any new worker to, by doing them, in the actual environment, with the actual equipment.
Hyundai, which owns Boston Dynamics, is also reportedly eyeing a factory in the Czech Republic as the next proving ground for Atlas in Europe.
Figure's humanoid robot walked into 30 homes it had never seen before (and did some work)
Figure AI's Figure 3 successfully navigated 30 previously unseen homes and achieved a 56% success rate on tasks it attempted inside them.
Think: cleaning up objects, folding stuff, moving things, making beds.
Let's be honest: 56% sounds low if you're comparing it to an average human. On the flip side, it sounds high if you understand what's actually being measured: a humanoid walking into an unfamiliar house, making sense of an environment it was never trained on, and completing real tasks in it.
Generalization — the ability to operate outside your training distribution — is the hardest problem in physical AI.
Figure isn't claiming the home robot is ready for your living room. But they're probably adding more real-world data on home tasks every day than every other humanoid robot company combined.
Faraday Future just launched nine new robots — including a $9,990 humanoid
Faraday Future is expanding far beyond cars.
The California company just launched nine new robot configurations across five models, spanning humanoids, quadrupeds, and other robotic forms.
One of the most interesting is the FF Master Mini, a compact humanoid starting at $9,990. It's aimed at programming education, open development, and robot sports.
But Faraday Future isn't betting on a single humanoid.
Its strategy is what it calls "one brain, multiple forms"... a broader robotics ecosystem where different machines share an embodied-AI platform. The new lineup ranges from relatively inexpensive smaller robots all the way to wheeled-legged systems priced at $137,900.
(Wheeled-legged quadrupeds have 4 legs, but each legs ends in a wheel. That makes them fast (and power efficient) on level surfaces, AND capable on terrain.)
The company is also pitching complete solutions for education, research, security, and inspection rather than simply selling robot hardware.
Human challenges a humanoid to a cage fight. The robot wins.
Over the weekend in San Francisco, influencer Frankie LaPenna stepped into a cage with EngineAI's T800 humanoid — a 6-foot robot built by a Shenzhen startup — in what was billed as the "first human versus robot fight," organized by REK, Robot Entertainment Kombat.
LaPenna landed some punches early. Then the robot knocked him to the canvas. He reportedly left with a hand or wrist injury.
REK publicly claimed LaPenna won, but the footage isn't so clear.
Important note: the T800 was remotely piloted by a human backstage, wearing a VR headset. The robot's onboard AI handled balance and posture stability. Every strike was a human decision.
It was a teleoperation match, not a fight against machine intelligence.
Number of the day: 7,000
That's how many humanoid robots the International Federation of Robotics says were sold worldwide in 2025, across industrial and professional service applications.
7,000.
That's dramatically lower than some of the numbers we've been hearing.
Omdia previously estimated about 13,300 humanoid robots shipped globally in 2025. A Chinese industry group put shipments from Chinese manufacturers alone at roughly 20,000.
So which number is right?
Maybe all of them ... because they're measuring different things.
Orders aren't shipments. Shipments aren't necessarily completed sales. And different research groups are using different definitions of what qualifies as a humanoid. IFR's new dataset excludes consumer, military, and medical robots, and this is the first year the organization has separately tracked humanoids at all.
We've seen the same problem in 2026. Counterpoint estimated more than 22,000 humanoids shipped globally in the first half of this year. Smart Analytics Global counted 19,100. Meanwhile, China's Humanoid Robot Scene Application Alliance said Chinese vendors alone shipped more than 30,000.
That's an enormous spread.
What the IFR number does reinforce is just how early this market still is. Most of the 7,000 robots it counted went to research, AI development, and pilot programs, not armies of humanoids doing productive work in factories.
The exact numbers are fuzzy.
The reality is: we're still very, very early.
Sources
Written by
I tell stories with data and build startups. I've raised angel and VC funds, led software companies and startups, and chronicled the rise of the mobile economy. Currently I write for Forbes, host several podcasts, and am starting up HumanoidMegaHub. I was TUNE's mobile economist and Singular's VP of Insights, built the research team at VentureBeat, and managed teams creating software for partners like Intel and Disney. In past lives I built websites, e-learning platforms, and mobile apps, and, a long time ago, was named to Folio's top 100 of the media industry's "most innovative entrepreneurs and market shaker-uppers.”