NEO's new 25-degree-of-freedom tendon-driven hands by 1X Technologies are the most capable robotic hands ever built, turning every joint into both a motor and a sensor.
Photo source:
neos-hands
A humanoid robot hand with force-controlled
joints is only meaningful if the capability it unlocks goes far beyond picking
and placing objects, and this is where NEO's hands demonstrate a genuine step
change. The 25 degrees of freedom are distributed the way human anatomy
distributes them, biased toward a thumb that genuinely opposes, enabling a
repertoire of grasps rather than a single mechanical grip. The result is a hand
that can assemble LEGO structures, pick individual screws and coins from a
wallet, spin and install light bulbs, use a screwdriver, rotate objects
in-hand, zip a jacket, sort grapes by color, pour tea from a kettle, catch a
squishy ball, plug in a USB-C charger, grab a wine glass, wipe surfaces, and
communicate via sign language.
That breadth of capability comes directly from
the combination of force transparency and tactile skin. High-resolution tactile
sensing across the fingertips and surfaces measures normal force, contact
location, and shear, the sideways force that tells the hand something is
beginning to slip. When an object starts sliding, the reflex catches it before
the slip completes, the same way a human hand tightens instinctively around a
glass that begins to fall. In addition, peak torques reach 3.5 Nm at the thumb
and 2.6 Nm at the finger joints, with wrist torque of 17.75 Nm, enabling strong
whole-hand grasps, tool use, lifting, carrying, opening doors, and pushing
loaded carts, all while retaining full dexterity at the small-object scale
where most real household tasks actually happen, with positioning accuracy of
plus or minus 0.2 millimeters.
The technical architecture of NEO's hands
reflects a specific philosophy: a hand that learns by touching everything must
be durable enough to keep touching everything. The motors live in the forearm,
where most human grip strength originates, pulling proprietary tendons through
the wrist. This keeps the hand itself lightweight while producing high forces
and running cool enough for continuous operation. The design integrates
in-house motors, custom electronics, embedded sensing, proprietary tendon systems,
compact transmissions, and hand-specific firmware in a single vertically
integrated system built and tested end-to-end in-house, from tendon materials
and soft polymers through to the tactile sensing stack.
That durability has been validated through
millions of test cycles across every subsystem, with wrist joints proven
reliable well beyond two million cycles under high loads and drive units tested
at extreme temperatures. The entire hand is sealed to IP68 with food-safe
materials, allowing NEO to work at the sink and wash its own hands after
completing household tasks, an industry first for a consumer humanoid.
Compliance closes the loop on safety too: low gear ratios combined with the
tendon drive allow external impacts to safely backdrive the fingers, meaning
the hands yield naturally when someone accidentally pushes against them rather
than resisting with damaging force. Hundreds of these hands have already come
off a scalable production line, with 1X targeting 10,000 hands produced in
2026, not as a manufacturing milestone but as a data strategy, since every hand
deployed in a real home runs experiments at scale and returns the training data
that embodied AI needs to keep learning.
The broader significance of NEO's hands extends
well beyond the specification sheet. A humanoid robot equipped with a
two-finger gripper exposes exactly three verbs to developers: pick, place,
push. Every application built on that platform is a composition of those three
verbs forever, executed without feedback. The ceiling is not in the software.
It is at the end of the arm. NEO's 25-degree-of-freedom, force-transparent,
tactile hands change that ceiling entirely. Every joint is simultaneously a
motor and a sensor. Every grasp returns pre-labeled data. Every probe returns a
measurement. This is the substrate learning-based manipulation has been waiting
for, and it means that future software improvements can unlock new capabilities
without ever redesigning the hardware again.
For a home robot, that last point matters more
than any dexterity demonstration. NEO is aimed at household use, targeted at
tasks like folding laundry, preparing meals, cleaning kitchens, and handling
fragile objects, in environments where safety, reliability, and consistency matter
as much as raw capability. These hands, shipping on every NEO unit, represent
the moment when the home robot promise stops depending on hardware
breakthroughs that haven't arrived yet, because the breakthrough just shipped.
Please subscribe to have unlimited access to our innovations.