Weve already been impressed by the astonishing balance and control of Unitrees G1 humanoid robot, with an earlier video showing it performing flips and other acrobatic moves with aplomb, and another showing it recovering at great speed after being pushed to the ground.

Now, researchers at the Beijing Academy of Artificial Intelligence (BAAI) have trained a Unitree G1 robot to pull a 1,400-kilogram car along a flat surface. The robot itself weighs just 35 kilograms and stands at 132 centimeters, so the feat does look rather remarkable. You can watch it in the clip below:

Researchers at Beijing Academy of Artificial Intelligence (BAAI) trained a Unitree G1 to pull a 1,400 kg car.

pic.twitter.com/VbWDAiDtNO

The Humanoid Hub (@TheHumanoidHub) October 28, 2025 In truth, the amount of force needed to move the car is relatively low as the vehicle is on a smooth, flat surface, with friction and rolling resistance very much on the low side.

Recommended Videos Whats most impressive here is the G1 robots ability to maintain its balance and steadiness while hauling the car along. Performing autonomously, watch how the robots feet and legs move rapidly on the slippery surface while also leaning back sharply to keep the car moving.

A robot with this level of balance and control would make it ideal for working among humans in somewhere like a warehouse environment, with little risk of it causing accidents through clumsiness.

It could also function well in search-and-rescue operations in debris-filled disaster sites where the terrain will likely be hard to navigate.

Related: Tesla close to something really tremendous with its humanoid robot, CEO Musk insists The G1 could also be deployed in other areas such as healthcare, performing tasks in hospitals or even homes where careful movement may be needed due to the limited space.

Unitree is one of a bunch of tech companies developing humanoid robots with a view to mass production. But challenges remain for developers of humanoid robots, particularly in areas like hand movement, so it could still be several years before we see their meaningful deployment in work settings and the like.