On April 19, 2026, the Honor/Monkey King team's fully autonomous humanoid robot "Lightning" won the Beijing E-Town Half-Marathon in a stunning 50:26 — beating the human world record by nearly 7 minutes and showcasing massive leaps in endurance, balance, and navigation.
Figure AI's Figure 03 demonstrates continuous unsupervised operation and full-body autonomy via Helix 02. 2026 roadmap includes factory deployments, robot-built-robot lines within 24 months, and home testing for complex, adaptive tasks in novel environments.
Figure AI has pushed Figure 03 into new territory with demos of 24/7 fully autonomous operation without human supervision, including overnight runs and real-world mobility like jogging outdoors at ~2 m/s. This builds on the January 2026 Helix 02 release, enabling true full-body vision-language-action control for unpredictable settings.
CEO Brett Adcock outlined aggressive 2026 goals: deploy on production lines this year, achieve robot-built robots in 24 months, deliver superhuman speed/precision hardware upgrades, and enable home robots for long-horizon tasks (e.g., full household chores) in completely unseen environments by year-end.
Figure 03 stands 5'8" (1.72 m), weighs 61 kg, runs 5 hours on a single charge, and handles 20 kg payloads. Redesigned hands, high-frame-rate vision, and soft/washable exterior prioritize safety and dexterity for home/factory use. It's engineered for Helix AI to learn from human demos and generalize across tasks like laundry, dishwashing, and navigation.
From folding laundry to running laps unsupervised, Figure 03 is shifting from prototype spectacle to scalable reality. If 2026 delivers on factory fleets and home pilots without constant babysitting, this could mark the point where humanoids stop being demos and start being daily tools.
BotQ manufacturing facility targets initial 12,000 units/year, scaling to 100,000 over four years. Recent pilots (e.g., BMW evaluations extending to Europe) test multifunctional applications, while home focus differentiates from pure industrial competitors.
Founded 2022, Sunnyvale, CA. Backed by NVIDIA, OpenAI, Microsoft, Bezos. Helix VLA model drives end-to-end autonomy. Team combines deep AI/robotics expertise for general-purpose deployment in homes and industry.
Production line deployments; robot-built lines by 2028; home pilots with adaptive long-horizon tasks; hardware upgrades for speed/precision. Vision: billions of humanoids in workplaces and homes.
While Atlas and Digit lead in automotive fleets, Figure 03's home emphasis and rapid autonomy gains position it as a frontrunner for consumer adoption. 2026 will test whether the AI-hardware combo can deliver reliable, unsupervised performance at scale.
Boston Dynamics unveiled the production version of Atlas at CES 2026 — a fully electric, enterprise-grade humanoid robot now entering manufacturing. All 2026 units are committed to Hyundai's Robotics Metaplant Application Center and Google DeepMind, with broader customer orders opening in 2027.
Boston Dynamics pulled the curtain back on the production version of Atlas at CES 2026 in Las Vegas — not a prototype, not a proof of concept, but a commercial robot now rolling off a line at the company's Boston headquarters. The reveal came during Hyundai Motor Group's global media day, where an electric prototype walked onto a stage in front of a live audience for the first time ever in public, before the sleek, blue production-ready version was unveiled behind a curtain.
The announcement carried weight beyond the spectacle. Every Atlas unit scheduled for 2026 is already spoken for — fleets are heading to Hyundai's Robotics Metaplant Application Center (RMAC) and Google DeepMind. Additional customers will be onboarded starting in early 2027, once Boston Dynamics has built confidence in fleet-scale operations.
Atlas is designed for the demands of industrial work: material handling, order fulfillment, parts sequencing, and machine tending. It operates autonomously, via teleoperator, or through a tablet steering interface — and crucially, it never needs to stop for a charge. The robot walks itself to a charging station and swaps its own battery in under three minutes.
The production Atlas stands 1.9 m (6.2 ft) tall, weighs 90 kg (198 lb), and reaches up to 2.3 m (7.5 ft). It lifts up to 50 kg (110 lb) instantaneously and 30 kg (66 lb) on a sustained basis, and operates across a temperature range of -4°F to 104°F. Its IP67 rating means it can be hosed down. The robot features human-scale hands with tactile sensing — a four-digit gripper with three fingers and an opposable thumb — engineered to handle the fine manipulation that automotive assembly demands. All limbs can be replaced in the field in under five minutes.
For years, Boston Dynamics was a studio that made remarkable films no one could buy. Atlas danced, did parkour, and amazed the internet — then disappeared back into a lab. At CES 2026, that era ended. The question was never whether Boston Dynamics could build the world's most capable robot. It was whether they could build one a factory could actually use. They have answered.
Alongside the hardware reveal, Boston Dynamics announced a new partnership with Google DeepMind to integrate Gemini Robotics foundation models directly into Atlas — giving the robot greater cognitive capability, adaptability, and the ability to generalize tasks across environments. DeepMind is receiving its own Atlas fleet to accelerate this work, making the collaboration as much a research deployment as a commercial one.
Once a single Atlas learns a new task, that skill replicates across the entire fleet instantly — a key differentiator in industrial environments where training time is a real cost.
Hyundai Motor Group is Boston Dynamics' majority shareholder and first customer. The RMAC opens in 2026 as a "data factory" for training Atlas. HMGMA in Bryan County, Georgia will receive Atlas by 2028. Hyundai Mobis is supplying actuators for Atlas, building a dedicated component supply chain.
Founded 1992, Waltham, Massachusetts. Over 30 years of robotics research. In 2025, deployed 500+ robots generating ~$130M in revenue from Spot and Stretch. Atlas named Best Robot at CES 2026 by the CNET Group. Majority owned by Hyundai Motor Group since 2020.
Hyundai plans to use Atlas for parts sequencing by 2028, extending to component assembly by 2030 — and eventually to repetitive, heavy, and complex operations across its global manufacturing network. The RMAC serves as a "data factory," building the training datasets that will teach Atlas the full range of tasks it will eventually perform at automotive volumes.
Boston Dynamics will open Atlas orders to additional early-adopter customers in early 2027, as the 2026 production run moves from pilot to proven deployment.
Agility Robotics and Toyota Motor Manufacturing Canada have signed a Robots-as-a-Service agreement to deploy seven Digit humanoid robots at the Woodstock, Ontario RAV4 plant — the first commercial humanoid deployment in Canadian automotive manufacturing.
Agility Robotics and Toyota Motor Manufacturing Canada (TMMC) have signed a commercial agreement to deploy seven Digit robots at the Woodstock, Ontario assembly plant — under a Robots-as-a-Service (RaaS) model.
The agreement follows a year-long evaluation at the same facility, run across three phases: development, proof-of-technology, and a live onsite pilot — with three Digit units involved throughout. TMMC, Toyota's largest manufacturing operation outside Japan, assessed the robots against real-world production standards before converting the engagement into a full commercial contract.
The seven Digit robots will perform one task: loading and unloading totes from automated tuggers that feed the assembly line. The scope is deliberate — Agility targets roles that are highly repetitive, physically taxing, and difficult to keep staffed.
TMMC's Woodstock plant — roughly 80 km southwest of Toronto — ran the pilot in a controlled area of the production floor, where engineers and workers observed Digit navigating an active automotive environment: forklifts, foot traffic, shifting layouts, and the constant tempo of a live supply chain.
What makes this agreement different from the wave of humanoid announcements flooding the industry is the word "commercial." Not pilot. Not trial. Not proof-of-concept. Toyota — a company that spent decades perfecting a manufacturing philosophy built on zero waste and zero error — has looked at Digit and decided it belongs on the same floor as its people. That is a more rigorous endorsement than any press release.
Agility and TMMC have committed to assessing further use cases beyond tote handling — exploring how automating the most physically demanding production-line tasks could reduce strain on workers.
Looking ahead, Agility is developing a next-generation Digit with a payload capacity of up to 50 lb (22.6 kg) and improved battery life, alongside ISO functional safety certification that would make Digit the first humanoid cleared to work cooperatively alongside people with no physical barriers — targeted for mid-to-late 2026.
Toyota Motor Manufacturing Canada operates plants in Cambridge and Woodstock, Ontario — Toyota's largest operation outside Japan. Over 8,500 employees. More than 535,000 vehicles in 2025, over 11 million since 1988.
Founded 2015, Salem, Oregon. Digit is the world's first commercially deployed humanoid. Customers include GXO Logistics, Schaeffler, Amazon, and Mercado Libre. Backed by Amazon, NVIDIA, Sony, and TDK.
As tariff pressure reshapes North American auto production, TMMC's investment in humanoid automation signals confidence in its long-term Ontario operations. Agility holds a growing lead in commercial traction: TMMC joins a roster of paying customers no competitor in the humanoid space has yet matched.
Global Momentum Accelerates: Honor’s Agile Demo at MWC, Tesla Optimus Gen 3 Production Timeline, BofA’s 3 Billion Forecast, Household Funding Surge & Beijing Robot Half-Marathon Trials
A journey through the most groundbreaking moments in humanoid robotics
When Honda pulled back the curtain on ASIMO at their Tokyo headquarters, the world witnessed something extraordinary. This wasn't just another robot—it was a 4-foot tall ambassador of the future, walking up stairs with an almost casual grace that made decades of robotics research suddenly feel worth it. Engineers had spent years perfecting the balance systems that let ASIMO navigate like we do, and watching it recognize faces and respond to voice commands felt like science fiction stepping into reality. The robot became Honda's technological pride, touring the world and inspiring a whole generation of researchers to push harder. What made ASIMO truly special wasn't just its technical specs, but how it made people believe that helpful humanoid robots in our homes and workplaces might actually happen someday.
Read more →Sony decided to tackle one of robotics' toughest challenges: making a humanoid robot actually run. QRIO was small, standing just two feet tall, but what it accomplished was huge. Running isn't like walking—both feet leave the ground, balance becomes critical, and one wrong calculation means a faceplant. Yet QRIO managed to hit speeds of 14 meters per minute, landing each step with precision that required incredibly sophisticated control algorithms. The robot could also recognize faces, respond to its name, and even dance, making it feel less like a machine and more like a companion. While Sony eventually discontinued the project, QRIO proved that dynamic movement was achievable and set the stage for the parkour-performing robots we'd see years later.
Read more →The Fukushima nuclear disaster had shown the world that some situations are just too dangerous for humans. DARPA and Boston Dynamics answered with ATLAS, a 6-foot, 330-pound humanoid built specifically for disaster response. This wasn't meant to be cute or friendly—ATLAS was designed to go where buildings have collapsed, where radiation levels are deadly, where rescue workers can't safely tread. With 28 hydraulically-actuated joints, it could navigate rubble, turn valves, use tools, and even drive vehicles. The robot's sensor suite gave it incredible situational awareness, letting it perceive its environment in 3D and make decisions on the fly. ATLAS represented a shift in thinking about humanoid robots—not as companions or entertainers, but as life-saving tools that could operate in our most catastrophic moments.
Read more →SoftBank took a completely different approach with Pepper, designing a robot specifically to understand and respond to human emotions. Standing 4 feet tall with large, expressive eyes and a tablet on its chest, Pepper was built to work alongside people in stores, banks, and hospitals. What made Pepper revolutionary was its emotional intelligence—sensors analyzed facial expressions, voice tone, and body language to gauge how someone was feeling and adjust its responses accordingly. Within months, Pepper was greeting customers in hundreds of locations across Japan, answering questions, providing directions, and even cracking jokes. Thousands were eventually deployed worldwide, proving that people were ready to interact with robots in everyday settings. Pepper showed that the future of humanoid robots wasn't just about physical capability, but about social and emotional connection.
Read more →The DARPA Robotics Challenge Finals in California brought together the world's best robotics teams for the ultimate test. Teams had to complete eight disaster-response tasks: driving a utility vehicle, getting out and walking across rubble, opening doors, using power tools, turning valves, and more. Team KAIST from South Korea took the top prize with their DRC-HUBO robot, which could switch between walking on two legs and rolling on wheels for efficiency. The competition wasn't just about winning—it was about proving what was possible. Robots fell over, got stuck, and struggled with tasks humans find trivial, but they also succeeded in ways that seemed impossible just years before. The challenge sparked innovations in autonomy, perception, and manipulation that continue to drive the field forward today.
Read more →Boston Dynamics dropped a video that made the world stop and stare. There was ATLAS, trudging through snow-covered woods, opening doors with handles, picking up boxes, and—in the moment that went viral—getting shoved by a researcher with a hockey stick and catching itself before falling. The internet collectively gasped. This wasn't a carefully choreographed demonstration on a clean lab floor; this was a robot operating in the messy, unpredictable real world. The door-opening sequence was particularly impressive, showing ATLAS turn a handle, pull the door, and hold it while walking through—a complex series of actions requiring precise force control and spatial awareness. The video racked up millions of views and sparked debates about robot rights (should we be pushing them?) and capabilities (what else can they do?). It was a watershed moment that showed humanoid robots were becoming genuinely capable machines.
Read more →Saudi Arabia made headlines worldwide by granting citizenship to Sophia, Hanson Robotics' humanoid AI robot. It was unprecedented—no country had ever given legal status to a non-human entity like this. Sophia, with her remarkably lifelike facial expressions and ability to hold conversations, had already become famous through TV appearances and conferences. But citizenship? That raised serious questions. What does it mean for a robot to be a citizen? Does Sophia have rights? Responsibilities? The decision was controversial, with critics calling it a publicity stunt and others seeing it as a forward-thinking recognition of AI's growing role in society. Regardless of where you stood, Sophia's citizenship forced the world to confront questions we'll increasingly face as AI and robotics advance. The stunt worked though—everyone was talking about robots, rights, and what personhood might mean in an age of intelligent machines.
Read more →Boston Dynamics released another jaw-dropping video, this time showing ATLAS doing parkour. Not walking. Not carefully stepping over obstacles. Parkour. The robot ran across platforms, jumped gaps, and landed backflips with a confidence that seemed almost showboating. Each movement required split-second decision-making, perfect balance, and incredible force control to stick the landings. The backflip alone was extraordinary—ATLAS had to generate enough rotational force, track its position mid-air, and time the landing perfectly. Engineers had developed new control algorithms that let the robot plan dynamic movements in real-time, adjusting on the fly to variations in the environment. Watching ATLAS move with such athletic ability was surreal; it was starting to move less like a machine and more like an athlete. The video went massively viral, showing that humanoid robots were rapidly approaching human-level agility in ways that seemed impossible just a few years earlier.
Read more →In what became one of 2020's most-watched videos, Boston Dynamics showed ATLAS, Spot, and Handle dancing in perfect synchronization to "Do You Love Me" by The Contours. It was mesmerizing. The robots twisted, jumped, spun, and grooved with moves that were simultaneously precise and playful. ATLAS threw in spins and even did the running man, while the dog-like Spot bounced along and Handle glided around on wheels. The choreography required incredible control—each robot had to coordinate its movements with the others while executing complex, dynamic motions. More than just a technical demonstration, the video showed that robotics could be joyful, creative, even entertaining. It humanized these machines in a way that technical specs never could. The video earned tens of millions of views and became a cultural moment, reminding everyone that even as robots become more capable, there's still room for whimsy and fun in how we think about them.
Read more →Elon Musk took the stage at Tesla AI Day and unveiled something nobody saw coming: Tesla Bot, later renamed Optimus. A 5'8" humanoid robot weighing 125 pounds, designed to handle tasks that are dangerous, repetitive, or just boring. Musk's pitch was characteristically ambitious—Optimus would leverage Tesla's massive advantage in AI, computer vision, and manufacturing to build a general-purpose humanoid robot at scale. Unlike many robotics companies building small numbers of research platforms, Tesla aimed to mass-produce millions of units at accessible prices. The plan was to use the same neural networks that power Tesla's self-driving cars to give Optimus the ability to navigate and manipulate objects in the real world. Critics were skeptical—building humanoid robots is notoriously hard—but Musk had a track record of achieving seemingly impossible things. The announcement instantly made Tesla a major player in humanoid robotics and signaled that the technology was moving from research labs into serious commercial development.
Read more →Just over a year after the announcement, Tesla wheeled out an actual working Optimus prototype at AI Day 2022. Admittedly, it was early-stage—the robot walked slowly, waved to the crowd, and demonstrated basic tasks like watering plants and moving boxes in a video. But the fact that Tesla had built anything at all in twelve months was impressive. The prototype showed off custom actuators, battery packs integrated into the torso, and hands with 11 degrees of freedom capable of delicate manipulation. Musk acknowledged they had a long way to go but projected confidence that Optimus could eventually be produced for under twenty thousand dollars. The demo was polarizing—some saw a promising start, others saw an overhyped concept. Regardless, it proved Tesla was serious about humanoid robots and willing to pour resources into development. The bigger message was clear: the era of general-purpose humanoid robots was arriving faster than most people expected.
Read more →Figure AI burst onto the scene with Figure 01, a sleek humanoid robot designed explicitly for commercial work. Unlike research platforms, Figure 01 was built with real-world deployment in mind—warehouses, manufacturing facilities, retail environments where labor shortages were becoming critical. The robot featured impressive dexterity with articulated hands that could manipulate a wide variety of objects, from boxes to tools to fragile items. Figure AI attracted serious attention, securing partnerships with major companies and raising significant funding from investors betting that humanoid robots were about to become practical reality. The company's approach was pragmatic: start with structured environments where tasks are well-defined, prove the robots can be useful and reliable, then expand capabilities over time. Figure 01 represented a new wave of robotics startups focused on commercialization rather than pure research. With labor markets tight across industries and warehouse automation already proven by companies like Amazon, the timing seemed right for humanoid robots to finally move from labs into workplaces at scale.
Read more →Select robots with checkboxes → Click "Compare Selected" for a clean side-by-side table.
All data verified as of April 2026. Scroll horizontally on mobile.
| Select | Robot | Manufacturer | Height | Weight | Payload | Speed (Walk) | Battery | Price (est.) | Status (April 2026) | Use Case |
|---|---|---|---|---|---|---|---|---|---|---|
| Tesla Optimus (Gen 2/3) | Tesla | 173 cm | 57 kg | 20 kg | 8 km/h | 6–8 hours | $20–30k (target) | Limited factory pilots, ramping production | General Purpose / Factory / Home | |
| Figure 03 | Figure AI | 170–173 cm | 60–61 kg | 20 kg | \~7 km/h | 5 hours | $50–100k (pilots) | BMW & other pilots; home trials | General / Home / Factory | |
| 1X NEO Gamma | 1X Technologies | 165–168 cm | 30 kg | 25 kg | 5 km/h | 4 hours | $20k or $499/mo | Pre-orders open, shipments 2026 | Home Assistant | |
| Agility Digit | Agility Robotics | 175 cm | 65 kg | 16 kg | 5.5 km/h | 8+ hours (swappable) | $200–250k or RaaS | Deployed in warehouses (Toyota, GXO) | Logistics / Warehouse | |
| Boston Dynamics Atlas (Electric) | Boston Dynamics | 188 cm (6'2") | 89 kg | 50 kg (instant) / 30 kg sustained | 9 km/h | 4 hours (self-swap) | Enterprise / Partners only | Industrial pilots (Hyundai) | Heavy Dynamic Tasks | |
| Unitree H1 | Unitree Robotics | 180 cm | 47 kg | 30 kg | 12 km/h (peak) | \~5–6 hours | $90k+ | Available for research | Research / Agile | |
| Unitree G1 | Unitree Robotics | 132 cm | 35 kg | 15 kg | 7 km/h | 2–4 hours | $13–16k | Widely available | Education / Research / Dev | |
| Apptronik Apollo | Apptronik | 173 cm | 73 kg | 25 kg | 5 km/h | 4 hours (swappable) | $50–100k (target) | Pilots ongoing | Industrial / Logistics | |
| Fourier GR-2 | Fourier Intelligence | 175 cm | 63 kg | \~20 kg (arm \~3 kg dexterous) | 5 km/h | 2 hours | $150k est. | Available / Mass production ramp | Rehab / Industrial / Research | |
| Ameca | Engineered Arts | 170 cm | 80 kg | Low | Stationary / Slow | Plugged in | $100k+ | Available | Social / Entertainment / Research |