CtrlK
| Figure 03 |
|---|
 |
| General information |
| Manufacturer |
| Country of origin |
| Year unveiled |
| Predecessor |
| Status |
| Target price |
| AI system |
| Website |
Figure 03 is a third-generation general-purpose humanoid robot developed by Figure AI, an American robotics company founded by Brett Adcock. Unveiled on October 8, 2025, the Figure 03 represents a significant redesign from its predecessor, the Figure 02, shifting the company's focus from purely industrial applications toward consumer and home environments. It was named one of TIME's Best Inventions of 2025 and appeared on the cover of the magazine's annual Best Inventions issue.[1][2]
The robot stands 1.68 meters tall, weighs 60 kilograms, and features 44 degrees of freedom across its body. It is powered by Helix, Figure AI's proprietary vision-language-action model (VLA), which enables the robot to perceive its environment, understand natural language commands, and execute complex manipulation tasks in real time. Figure 03 is designed for mass production at BotQ, Figure AI's in-house manufacturing facility in San Jose, California, with an initial capacity of 12,000 units per year.[3]
Figure AI was founded in 2022 by Brett Adcock, a serial entrepreneur previously known for founding Archer Aviation (an electric vertical takeoff and landing aircraft company) and Vettery (a hiring marketplace). The company set out to build commercially viable general-purpose humanoid robots capable of operating in environments designed for humans.[4]
The company's first prototype, Figure 01, emerged from stealth in March 2023. The Figure 01 achieved dynamic bipedal walking in under a year of development, which Figure AI described as one of the fastest turnarounds in humanoid robotics history. Standing 1.68 meters tall and weighing 60 kilograms, Figure 01 demonstrated basic warehouse tasks including walking and box manipulation. It used external cabling and was primarily a research platform rather than a production-ready system.[5]
In August 2024, Figure AI unveiled the Figure 02, its second-generation robot designed for industrial deployment. The Figure 02 featured 41 degrees of freedom (including 16 per hand), five-fingered hands, a 25-kilogram payload capacity, six RGB cameras, integrated internal cabling (replacing the external wiring of Figure 01), and three times the computing power of its predecessor. It was the first Figure robot to run an onboard vision-language-action model.[6]
A pivotal moment in Figure AI's development came through its commercial partnership with BMW Manufacturing, announced in January 2024. Figure 02 robots were deployed at BMW's plant in Spartanburg, South Carolina, where they participated in automotive production operations. Over an 11-month deployment, the Figure 02 fleet accumulated more than 1,250 operational hours, loaded over 90,000 sheet-metal parts, and contributed to the production of more than 30,000 BMW X3 vehicles.[7]
The deployment generated critical data that directly shaped the Figure 03 design. The forearm emerged as the most common hardware failure point during factory operations. For Figure 03, the engineering team completely re-architected the wrist electronics, eliminating both the distribution board and dynamic cabling. Each wrist's motor controller now communicates directly with the main computer, reducing complexity, improving reliability, and simplifying thermal management.[7]
In 2026, BMW expanded its humanoid robotics program to Europe, with a pilot deployment at its Leipzig plant in Germany, establishing a "Center of Competence for Physical AI in Production" to accelerate global integration of AI and robotics.[8]
Figure AI initially collaborated with OpenAI beginning in 2024, integrating OpenAI's large language models to enable speech-to-speech communication on the Figure 02. OpenAI also participated as an investor in Figure AI's February 2024 funding round.[9]
However, on February 5, 2025, Figure AI ended the partnership. CEO Brett Adcock cited two primary reasons: first, OpenAI announced plans to resurrect its own robotics program, making the two companies potential competitors; second, Figure AI concluded that building effective embodied AI required vertical integration of the robot's AI systems. Adcock stated, "We found that to solve embodied AI at scale in the real world, you have to vertically integrate robot AI." Following the split, Figure AI developed its Helix model entirely in-house.[10]
Figure 03 maintains a humanlike form factor at 1.68 meters (5 feet 6 inches) tall, the same height as its predecessors. It weighs 60 kilograms (132 pounds), making it 9% lighter than the Figure 02 (which weighed approximately 70 kilograms). The reduced mass, combined with a more compact overall volume, enables easier navigation through standard doorways, hallways, and household environments.[3]
A defining design change in Figure 03 is its exterior. Rather than the exposed hard plastic and metal shells used on the Figure 01 and 02, the Figure 03 is covered in soft, washable textiles layered over multi-density foam padding. This covering eliminates pinch points and reduces the risk of injury during close-proximity interaction with people. The soft goods are tool-free removable and fully washable, and they can be customized with different garments, including options made from cut-resistant and durable materials.[3]
| Category | Specification | Value |
|---|---|---|
| Physical | Height | 1.68 m (5 ft 6 in) |
| Physical | Weight | 60 kg (132 lbs) |
| Physical | Exterior | Soft washable textiles with multi-density foam |
| Mobility | Total degrees of freedom | 44 |
| Mobility | DOF per hand | 16 |
| Mobility | Max walking speed | 1.2 m/s (4.3 km/h) |
| Manipulation | Payload capacity | 20 kg |
| Manipulation | Fingers per hand | 5 |
| Power | Battery | F.03 (2.3 kWh capacity) |
| Power | Runtime | ~5 hours at peak performance |
| Power | Charging | 2 kW wireless inductive (via foot coils) |
| Sensors | Cameras | 6 main cameras + 2 palm cameras (8 total) |
| Sensors | Tactile | Custom fingertip sensors (3-gram sensitivity) |
| Sensors | Data offload | 10 Gbps mmWave |
| Actuators | Speed improvement | 2x faster than Figure 02 |
| Actuators | Torque density | Improved Nm/kg over Figure 02 |
| AI | Model | Helix VLA (System 1 + System 2) |
| AI | Compute | Dual embedded low-power GPUs |
| Safety | Battery certifications | UN38.3, UL2271 (in process) |
Figure 03 features a completely redesigned sensory suite purpose-built to enable the Helix AI system. The camera architecture delivers twice the frame rate, one-quarter the latency, and approximately 60% wider field of view per camera compared to the Figure 02's vision system. The robot carries six main cameras for spatial perception and object recognition, plus an embedded camera in each palm for close-range grasping of occluded objects, bringing the total to eight cameras.[3]
The audio system was also upgraded substantially. The speaker is twice the size and nearly four times more powerful than the Figure 02's speaker, enabling clearer voice communication. Microphone positioning was improved for better speech recognition clarity in noisy home environments.[3]
Each of Figure 03's hands features 16 degrees of freedom and first-generation proprietary tactile sensors integrated into every fingertip. These custom-developed sensors can detect forces as small as three grams of pressure, which is roughly the weight of a paperclip resting on a finger. This level of sensitivity enables the robot to distinguish between a secure grip and an impending slip before it occurs, allowing fine-grained, dexterous control over fragile, irregular, or moving objects.[3]
The fingertips themselves were redesigned with softer, more adaptive surfaces that increase the contact area during grasps. This combination of tactile feedback and compliant fingertip geometry allows Figure 03 to handle delicate items such as eggs, glassware, and fabrics without damaging them.[3]
Figure 03 is powered by the F.03 battery, a 2.3-kilowatt-hour pack that delivers approximately five hours of runtime at peak performance. The battery represents a 94% increase in energy density compared to the first-generation F.01 battery and a 78% reduction in cost compared to the F.02 battery. Unlike the F.01, which used a bulky external backpack design, the F.03 battery is integrated directly into the robot's torso.[11]
The F.03 battery is the first humanoid robot battery in the process of being certified to both UN38.3 and UL2271 safety standards. It underwent 23 primary tests covering mechanical stress, thermal cycling, and electromagnetic interference. Safety features include a custom Battery Management System (BMS) with sensors, switches, and fuses that prevent overcharge, overdischarge, and overtemperature conditions. Cell-to-cell wirebond interconnect geometry is tuned to act as a fusible element, providing an additional layer of protection.[11]
The robot supports wireless inductive charging through coils embedded in its feet. Rather than requiring manual cable connections, Figure 03 can step onto a wireless charging stand and recharge at 2 kilowatts with active forced-convection cooling. The charging system also supports wireless data offload via 10 Gbps mmWave, enabling fleet-wide learning without physical connectivity.[3][11]
The Figure 03's actuators are capable of twice the speed of those in the Figure 02, with improved torque density (measured in Newton-meters per kilogram). These improvements enable picking and positioning speeds comparable to human workers. Moving from Figure 02 to Figure 03 required the mechanical and electrical engineering teams to redesign nearly every component with manufacturability and cost in mind, aggressively reducing part count, assembly steps, and any components that were not strictly necessary.[3]
Helix is Figure AI's proprietary vision-language-action model (VLA) that controls Figure 03. Developed entirely in-house following the end of the OpenAI partnership, Helix unifies perception, language understanding, and learned motor control into a single model. It is the first VLA demonstrated to output high-rate continuous control of an entire humanoid upper body, including wrists, torso, head, and individual fingers.[12]
Helix uses a dual-system architecture inspired by cognitive science's "System 1, System 2" framework:
Gradient backpropagation flows between S1 and S2 through latent vectors, enabling end-to-end training from raw pixels and text to continuous robot actions.[12]
Helix was trained on approximately 500 hours of high-quality teleoperated demonstrations collected across multiple robots and operators. This dataset is substantially smaller than those used by many competing VLA approaches. An auto-labeling vision-language model generated hindsight instructions from video clips, enriching the training data without additional human annotation effort.[12]
The model uses a single unified set of weights for all behaviors, including picking, placing, drawer operation, refrigerator use, and multi-robot handovers, without task-specific fine-tuning. Key capabilities include:
Figure AI has positioned Figure 03 as the company's first robot designed primarily for home environments. Demonstrations released alongside the October 2025 unveiling showed the robot performing a range of domestic tasks in simulated home settings. These included folding towels and clothing, loading and unloading dishwashers, clearing and wiping tables, sorting and loading laundry, watering plants, serving tea, and navigating between rooms while responding to voice commands.[13][14]
The robot's ability to perform these tasks relies on the combination of Helix's zero-shot generalization (allowing it to handle novel household objects), the tactile sensors (enabling gentle handling of fragile items), and the palm cameras (allowing manipulation of objects that are partially hidden or in tight spaces such as dishwasher racks).[3]
CEO Brett Adcock has stated that select home deployments are planned for 2026, with broader consumer availability expected in the 2026 to 2027 timeframe. Initial deployments will be limited to select partner households, and the robot will not be available for general consumer purchase until availability expands.[15]
To support mass production of Figure 03, Figure AI developed BotQ, a dedicated high-volume humanoid manufacturing facility located on the company's campus in San Jose, California. BotQ represents Figure AI's decision to bring humanoid robot manufacturing fully in-house, controlling the build process, quality assurance, and supply chain rather than relying on contract manufacturers.[16]
The first-generation BotQ production line has an initial capacity of up to 12,000 humanoid robots per year. Figure AI's goal is to produce a total of 100,000 robots over the following four years. The facility employs die-casting, injection molding, and stamping processes, and vertically integrates the production of actuators, batteries, sensors, and electronics. Figure AI also developed an internal Manufacturing Execution System (MES) to manage production workflows.[16]
A major achievement in the Figure 02 to Figure 03 transition was a 90% reduction in component costs. The engineering team redesigned nearly every component with manufacturability in mind, shifting from complex machining processes to low-cost, high-volume tooled processes. This cost reduction is central to Figure AI's goal of eventually offering the robot at a consumer-accessible price point below $20,000.[3][16]
Figure AI's development has been supported by several major funding rounds:
| Round | Date | Amount | Valuation | Key investors |
|---|---|---|---|---|
| Seed | May 2023 | $70 million | Not disclosed | Brett Adcock ($20M personal), Parkway Venture Capital |
| Series B | February 2024 | $675 million | $2.6 billion | Jeff Bezos, Microsoft, NVIDIA, Intel, Amazon, OpenAI |
| Series C | September 2025 | $1+ billion | $39 billion | Parkway Venture Capital, Brookfield, NVIDIA, Macquarie, Intel Capital, Qualcomm, Salesforce, T-Mobile, LG |
The Series C round, announced on September 16, 2025, represented a roughly 15-fold increase in valuation from the February 2024 round just 18 months earlier. Total funding raised across all rounds exceeds $1.9 billion, making Figure AI one of the most well-capitalized pure-play humanoid robotics companies in the world.[17][18]
Figure 03 enters an increasingly crowded humanoid robot market. Key competitors include:
| Company | Robot | Focus | Notable features |
|---|---|---|---|
| Tesla | Optimus Gen 3 | Industrial, consumer | Manufacturing scale, $20,000-$30,000 target price |
| 1X Technologies | NEO | Home, consumer | ~$20,000 price point, lightweight design |
| Agility Robotics | Digit | Warehouse, logistics | 10,000-unit/year factory, deployed in Amazon facilities |
| Unitree | G1 / R1 | Consumer, education | R1 priced at $5,900, ultra-lightweight |
| Boston Dynamics | Atlas (electric) | Research, industrial | Advanced mobility and acrobatics |
| Apptronik | Apollo | Industrial, logistics | NASA heritage, modular design |
| Sanctuary AI | Phoenix | General purpose | Carbon (proprietary AI), teleoperation-first approach |
The global humanoid robot market was valued at approximately $4.89 billion in 2025 and is projected to grow to $165 billion by 2034, reflecting a compound annual growth rate of over 50%. Chinese manufacturers, particularly Unitree and AgiBot, led global shipments in 2025 with over 10,000 robots combined, while U.S. companies including Tesla, Figure AI, and Agility Robotics each shipped roughly 150 units, underscoring the early stage of Western market development.[19]
Figure 03 differentiates itself through its home-first design philosophy (soft textiles, wireless charging, safety-focused exterior), its proprietary vertically integrated AI stack (Helix), and its in-house manufacturing capability (BotQ). The combination positions Figure AI to compete on both the consumer and commercial fronts as the market scales.
Figure 03 was named one of TIME's Best Inventions of 2025 and appeared on the cover of the magazine's annual Best Inventions issue, accompanied by a cover story by journalist Billy Perrigo profiling Figure AI's ambitions to create a mass-producible home humanoid.[1]
In March 2026, Figure 03 appeared at a White House event hosted by First Lady Melania Trump. The two-day summit, titled "Fostering the Future Together," focused on AI in education and featured diplomats from 45 nations. The robot's appearance generated significant global media coverage and public discussion about the role of humanoid robots in education and daily life.[20]
| Feature | Figure 01 (2023) | Figure 02 (2024) | Figure 03 (2025) |
|---|---|---|---|
| Height | 1.68 m | 1.68 m | 1.68 m |
| Weight | 60 kg | ~70 kg | 60 kg |
| Total DOF | N/A | 41 | 44 |
| Hand DOF | N/A | 16 | 16 |
| Payload | 20 kg | 25 kg | 20 kg |
| Battery runtime | ~5 hours | ~5 hours | ~5 hours (F.03, 2.3 kWh) |
| Cameras | Cameras + LiDAR | 6 RGB | 8 (6 main + 2 palm) |
| Tactile sensors | Basic | Force sensors | 3-gram fingertip sensors |
| Charging | Wired | Wired | Wireless inductive (2 kW) |
| Exterior | Hard shell, external cabling | Hard shell, internal cabling | Soft washable textiles + foam |
| AI system | None (teleoperated) | Onboard VLA + OpenAI LLM | Helix VLA (in-house) |
| Compute | N/A | 3x Figure 01 | Dual embedded GPUs |
| Target use | Research / warehouse | Industrial (BMW) | Home + commercial |