Figure 03

A humanoid designed for two settings

Figure introduced Figure 03, its third-generation humanoid robot, on 9 October 2025 ^[1]. The company said the machine was designed around Helix, its vision-language-action AI, and aimed at two settings at once: the home and mass manufacturing ^[1]. A vision-language-action model takes camera images and language instructions as input and produces motor actions as output, so the body has to supply the sensing and the strength that the model assumes.

A new camera architecture and a sense of touch

Figure describes a new camera architecture across the robot: each camera delivers twice the frame rate of the earlier design, with one-quarter of the latency and a 60% wider field of view ^[1]. Wider views and faster frames matter for a machine that has to notice a person moving nearby or an object slipping out of a grasp.

Each hand also carries an embedded palm camera, which provides close-range visual feedback when the main cameras are occluded — by the hand's own body, for example, or by the object being held ^[1]. Figure says the fingertip sensors can detect forces as small as three grams of pressure ^[1]. That level of sensitivity is what separates a gripper that can pick up a sponge or a piece of clothing from one that can only handle rigid parts.

Body, power and charging

Figure 03 has 9% less mass and significantly less volume than Figure 02, and is covered in washable, tool-free removable soft textiles and multi-density foam ^[1]. Removable, washable covers are a practical requirement for a machine that works in a house; foam padding changes how an accidental contact with a person feels.

Its battery has achieved certification to the UN38.3 standard ^[1], the transport-safety test applied to lithium batteries. Charging coils built into the feet allow wireless inductive charging at 2 kW, so the robot can take on energy by standing on a charging pad rather than by being plugged in by a person ^[1]. The robot also supports 10 Gbps mmWave data offload ^[1], a high-bandwidth wireless link for moving recorded data off the machine.

Figure says its actuators can run at 2x faster speeds with improved torque density ^[1] — more force per unit of actuator mass, and quicker motion from the same hardware.

Building them in volume

Alongside the robot, Figure describes its manufacturing plans: the first-generation line at BotQ can produce up to 12,000 humanoid robots per year, with a goal of 100,000 robots over the following four years ^[1]. Designing for a line rather than for a single prototype explains several choices on Figure 03, such as coverings that snap off without tools and a body with less mass and volume to move and assemble.

Published specifications

Figure's product page lists a height of 5 ft 8 in, a payload of 20 kg, a weight of 61 kg, a runtime of 5 hr, a speed of 1.2 m/s and an electric system ^[2]. The same page says the robot handles household tasks such as laundry, cleaning and doing dishes autonomously ^[2].

Where it sits

Figure 03 is the third generation of Figure's humanoid, and its announcement ties three things together: an AI model the hardware is built around, a body shaped by the constraints of a home, and a production line meant to build the robot in five figures per year ^[1]. The specifications read like those of a product rather than a demonstration: a stated payload, a stated runtime, a standard battery certification, and a charging method that does not need a person ^[2][1]. Whether the household tasks listed on the product page are performed reliably in real homes is a question the announcement itself does not answer; what it documents is the design intent and the numbers Figure attaches to it.