1X Neo Revealed Most Advanced AI Robot Hardware (25-DOF Upgrade)

[00:00:01]The race to build a humanoid robot that can work inside factories and handle chores at home is moving faster than ever, and every robotics company is trying to reach that goal in a different way. Unitree is focusing on movement by teaching its robots to walk across difficult ground, run at high speeds, recover their balance, and even perform live on stage. Boston Dynamics is building Atlas around strength and control, showing the robot lifting and carrying heavy objects like a mini fridge. UBTECH is taking another path with its U1 Ultra robots, which can blink, [music] track faces, hold surprisingly natural conversations, and work in roles such as reception and customer service. These companies may be following different paths, but they are all moving toward the same goal, building a humanoid that can perform useful work in factories, workplaces, [music] and eventually our homes. A robot's body movement matters.

[00:00:59]So does how much it can lift and how clearly it can see. But none of these abilities are enough if the robot cannot safely pick up tools, handle fragile objects, fold clothes, prepare food, or complete the countless physical tasks people perform every day.

[00:01:16]For that, a humanoid needs one thing above everything else, hands.

[00:01:21]The human hand is one of the most complex and important parts of the body.

[00:01:26]It can lift something heavy, hold something fragile, feel when an object begins to slip, and adjust its grip without us even thinking about it. Until now, engineers have usually been forced to choose between strength, precision, touch, safety, and durability.

[00:01:44]Very few designs have been able to deliver all of these abilities together.

[00:01:48]But one company believes it has finally figured it out. 1X has just revealed an entirely new hand for its Neo home robot. And it could remove one of the biggest hardware limits stopping humanoids from performing real work. It has 25 powered degrees of freedom, tendon-driven fingers, force-controlled joints, and tactile skin that can detect pressure and slipping. These are not experimental parts built for one demonstration. 1X says the new hands will ship on every Neo.

[00:02:21]To understand why this reveal is important, look at what other leading humanoids use. Atlas has powerful 7-degree of freedom grippers with tactile sensors and palm cameras. They can lift heavy objects and still perform smaller precision grasps. Figure 03 uses five-fingered hands with sensitive fingertips and cameras inside the palms.

[00:02:43]Unitree G1 is commonly shown with an optional three-finger hand offering 7 degrees of freedom, while Unitree's newer five-fingered Dex 5-1 reaches 20.

[00:02:55]Each design is improving, but engineers are still balancing human-like movement against strength, reliability, cost, and production. Neo's new hand is another major step in that race.

[00:03:07]Let's see what they can actually do. In 1X's demonstration, Neo builds LEGO structures and picks individual screws or coins from flat surfaces. The robot spins and installs a light bulb, uses a screwdriver, zips a jacket, and rotates objects inside its palm. It sorts grapes, pours tea from a kettle, catches a soft ball, plugs in a USB-C cable, holds a wine glass, cleans a surface, and forms signs used in sign language.

[00:03:36]These tasks require very different combinations of movement, pressure, balance, and finger placement. The hand is driven by tendons, so motors inside Neo's forearm pull thin cables connected to the fingers. This is closer to the basic structure of a human arm and hand, where muscles in the forearm pull tendons to move the fingers. Keeping the motors inside the forearm also allows the hand itself to remain lighter, while still producing useful force.

[00:04:05]The design has 25° of freedom. 22 are spread across the fingers and palm, while three control the wrist. Every one of those movements is powered and force-controlled. The fingers can therefore move separately, instead of opening and closing together like one stiff claw. For comparison, a common model gives the human hand and wrist around 27° of freedom. That makes Neo's total [music] surprisingly close on paper. However, the number alone does not prove that the hand is equal to ours. Human dexterity also comes from the shape of our joints, soft tissue, thousands of sensory receptors, muscle control, [music] and a brain that has spent years learning how to use them.

[00:04:49]Still, 25 independently controlled movements give Neo [music] far more options than a basic gripper. The thumb can move against the other fingers. The palm can change shape around an object.

[00:05:00][music] Individual fingers can adjust their positions, while the robot continues holding something. This is not simply a stronger claw attached to the end of an arm, it is a complete [music] hand designed for everyday work. The biggest breakthrough is something 1X calls force transparency. Most robot hands can be told to move toward a set position, but they cannot clearly feel what happens through their own joints after touching something. Many robot hands use gear ratios of 100 to 1 or even 200 to 1. Those gears can produce strong movement, but friction inside the transmission blocks much of the contact force before it travels back to [music] the motor. The robot may know where its finger was commanded to go, yet receive very little information about what stopped it. Engineers can add cameras or separate pressure sensors, but the joint itself remains mostly none. Neo uses much lower gear ratios of around 5:1 to 15:1. Its motors pull the fingers through a tendon system that allows the joints [music] to move backward when an outside force pushes against them. Press one of Neo's fingers and it gives way while measuring how hard it was pushed.

[00:06:07]Each joint can produce movement and collect information through the same physical path. The hand also has proprioception. [music] This means Neo knows the position of its fingers without needing to look directly at them. Humans use the same basic ability when we touch two fingertips together with our eyes closed. People do not understand objects through vision alone. We press something to test its hardness. We slide a finger [music] across it to feel texture. We squeeze it to learn how easily it bends. We change our grip when an object begins to move.

[00:06:40]Neo can now collect similar physical information through its joints and tactile Do you think 25 powered movements bring Neo truly close to a human [music] hand or is 1X making that claim too early?

[00:06:54]Share your thoughts in the comments. The hand is also designed to combine delicate control with useful strength.

[00:07:00]The thumb joint produces peak torque of around 3.5 Newton meters while the main finger joints reach around 2.6 Newton meters. The fingertips can produce up to 45 Newtons of force and the wrist delivers 17.75 Newton meters of torque. This allows Neo to create strong whole hand grips, carry objects, use tools, open doors, and push loaded carts. At the same time, 1X reports positioning accuracy of around 2/10 of a millimeter. That combination allows the same hand to lift a kettlebell and then carefully pick up a coin or small screw. The surfaces of the fingers contain high resolution tactile sensors. They measure normal pressure, the location of contact, and shear.

[00:07:47]Shear is the sideways force that appears when an object starts moving across the skin. Suppose a glass begins slipping from Neo's grip. The shear sensors can detect that movement before the glass completely slides away. The robot can then adjust its fingers and tighten the grip. Cameras may struggle when an object is small, transparent, soft, or partly hidden by the hand. Touch provides information that vision may miss. The hands are also sealed to an IP68 rating and made with food-safe materials. Neo can work near a sink and wash its hands after preparing or handling food. This is especially important for a robot intended to operate in kitchens, bathrooms, and other parts of a home where water and dirt are unavoidable. Durability may be even more important than one impressive demonstration. A home robot could touch thousands of objects each day. Its fingers will hit furniture, become trapped inside drawers, carry rough objects, and repeat the same movements over and over. Now, let's look at the robot carrying this hardware. Neo is 1X's humanoid for the home. Developed as a consumer product rather than only as a research robot or factory worker. CEO Bernt Burnik believes homes may provide better training data than warehouses because home tasks are more varied and less predictable. A robot inside a house must deal with different rooms, objects, people, and routines. That makes the home difficult, but it can also teach Neo a wider range of skills. Neo stands about 5 ft 6 in tall and weighs roughly 66 lb. Its soft body sits beneath a knitted suit, which helps it feel less threatening around people. The robot operates at around 22 dB, making it quieter than many refrigerators. Even with its light frame, Neo is designed to lift more than 150 lb. Neo costs $20,000, while a monthly option is priced at $499.

[00:09:50]Customers can place a refundable deposit, and more than 10,000 reservations have reportedly been made.

[00:09:56]US deliveries are planned through 2026, with other markets expected in 2027. 1X has opened a US factory and is targeting 10,000 robots this year. However, Neo is not fully autonomous yet. Early units will still depend on human-in-the-loop operation for harder tasks. A trained operator can guide the robot through a headset while Neo collects data from the work. Its autonomy has been described as roughly 60 to 70% In March 2026, 1X also showed a world model that allows Neo to learn by watching video instead of relying only on fixed instructions. This means the hardware can do more than the software currently knows how to control.

[00:10:39]The hands already have the movement, strength, and touch needed for many home tasks, but the AI still has to learn how to use those abilities reliably.

[00:10:48]Over-the-air updates could add new skills after delivery. Tesla and Figure are chasing the same home market, but every company faces the same challenge because laundry, dishes, cooking, cleaning, >> [music] >> and organizing all depend on what the hands can safely handle. The bigger idea is simple. A humanoid robot can walk through a house and understand what it sees, but its hands decide whether it can actually help. By combining strength, precise movement, touch sensing, and a design built for everyday environments, 1X has removed one of the biggest limits holding Neo back. 1X's components and complete finger assemblies have survived millions of testing cycles. The drive units were tested at extreme temperatures, while wrist joints continued operating beyond 2 million cycles under heavy loads. The low gearing also improves safety. When Neo's fingers hit something, the joints can move backward rather than remaining dangerously rigid. Slow-motion tests show the fingers yielding when slapped, hit with a hammer, trapped inside a closing drawer, or pushed into another object. This passive compliance is important around people, children, pets, and fragile items. Safety does not depend entirely on the AI noticing a collision before it happens. Some protection comes directly from the physical design of the hand. These are also not one-off parts made for a reveal video. Hundreds of the new hands have already come off a dedicated production line. 1X says the line has enough capacity to produce 10,000 hands during 2026.

[00:12:23]Each unit is manufactured and tested in-house, including the motors, tendons, electronics, soft materials, and tactile sensing.

[00:12:33]The software still needs time to catch up, but the hardware is now ready for skills that were difficult to imagine only a few years ago. Subscribe if you want to follow how this technology develops, because Neo is not the only robot pushing closer to useful human-level control. The next breakthrough may come from a completely different design. BMW just gave a humanoid robot one of the hardest jobs in its American factory. Figure 03 has arrived at BMW Group Plant Spartanburg in South Carolina, and the assignment is bigger than before. The deployment follows a successful pilot with the earlier Figure 02 robot inside the same plant. Figure 02 supported production of more than 30,000 BMW X3 vehicles over roughly 10 months. Working in the body shop, Figure 02 inserted sheet metal parts for the welding process. That task demands high speed, precision, and repeatability across every single shift.

[00:13:33]The pilot-proved humanoid robots can perform real production work safely alongside employees. Figure founder Brett Adcock says the deployment showed humanoids are no longer lab experiments.

[00:13:46]Now, Figure 03 moves into Hall 52, where BMW assembles the X3 and in the future, the electrified iX5. The new job is called sequencing and it sits inside the logistics operation. Parts arrive at the plant in large containers, completely unsorted. Figure 03 picks up individual components and sorts them into a sequencing trolley in production order.

[00:14:09]The trolley then moves to a collection point for onward transport. An automated tugger train or smart transport robot carries the parts to the assembly line.

[00:14:19]Employees receive exactly the right part at exactly the right moment. BMW calls this just-in-sequence delivery and it happens constantly in automotive logistics. Sequencing sounds simple, but it is one of the toughest jobs in a modern factory. Parts do not arrive in perfect positions every time. Components shift, rotate, or sit partially hidden inside their containers. That means hard-coded motions fail because no two picks look exactly alike. Figure 03 must perceive each scene and make small corrections on the fly. The robot grasps parts with both hands while adjusting foot placement and shifting balance. The technology making this possible is Helix 02, Figure's pixels to actions vision language action model. Helix 02 coordinates the hands, arms, torso, and feet [music] of Figure 03 as one system.

[00:15:12]In demonstrations, Figure 03 places thin-walled parts with precision, [music] then pulls a heavy metal cart on caster wheels. Fine control and full-body force in a single workflow.

[00:15:25]That combination is structurally infeasible [music] for fixed automation or a six-axis robotic arm. Figure 03 also brings hardware upgrades designed for the factory floor. Soft exterior components improve safety when working around people. Wireless charging keeps the robot [music] available longer without manual intervention. Audio functions enable [music] speech-to-speech communication with human co-workers. Improved hands carry tactile sensors and palm cameras for greater precision [music] and dexterity.

[00:15:56]BMW frames the project as part of a broader physical AI strategy, connecting digital intelligence with real machines.

[00:16:05]The company says humanoid [music] robots complement existing automation rather than replace employees. The targets are monotonous, ergonomically demanding, or safety-critical tasks. The goal is protecting workers while improving workplaces. Figure is expanding beyond BMW at the same time.

[00:16:24]The company signed a commercial agreement with Catalyst Brands, the retail group behind J.C. Penney, Aeropostale, and Brooks Brothers. Figure humanoids will enter Catalyst's distribution and logistics [music] network, starting at a center in Reno, Nevada. The deal also marks the first commercial bridge between two companies backed by investor Brookfield. All of this is landing while the robotics world waits [music] on Tesla. Elon Musk has said the Optimus Gen 3 design is now finalized. The robot [music] is already walking, and the official reveal is coming soon with production targeted around summer 2026, bigger scaling expected into 2027, [music] and a possible price range of $20,000 to $30,000 per robot. But while everyone waits for Optimus Gen 3 to be shown properly, Figure is already moving past [music] the robot that just went 10 rounds with a human. On May 13th, 2026, Adcock announced that Figure 04 had cleared its critical design review, is in full design lock, and has parts already shipping. If you enjoyed this video, make sure to subscribe to AI Nexus [music] for more AI and robotics updates. This was our July update on 1X Neo, but it is only one part of a much bigger shift. Click the video on your screen to see how far the entire AI robotics industry has progressed [music] over the past 6 months.