Figure’s Humanoid Robot Factory Just Hit a Major Production Milestone

Humanoids are no longer a boutique luxury. At Figure, they’re now on the clock.

The California-based robotics firm has announced a major production milestone for its Figure 03 model. In just four months, the company has transformed its “BotQ” facility from a prototype shop into a buzzing factory that now pumps out a new robot every hour.

The leap from building a few experimental machines to running a high-volume factory is often where tech startups stumble. Figure, however, claims to have accelerated past that hurdle. According to a recent company report, the team has achieved a 24x improvement in throughput in less than 120 days.

“The Figure hardware and manufacturing teams have transformed BotQ from a blueprint into a high-output environment, delivering over 350 of our third generation humanoid robots and increasing our production rate from 1 Figure 03 per day to 1 per hour – a 24x throughput improvement in under 120 days,” Figure wrote in a blog post.

This ramp-up is powered by a custom manufacturing system that connects over 150 workstations. It isn’t just about speed, either; the company says it has tightened its quality control, qualifying hundreds of suppliers and setting up 50 different inspection points to ensure the robots aren’t just fast, but reliable.

Robots that work out before they ship

Building a humanoid is one thing, but making sure it doesn’t break the moment it touches a customer’s floor is another. Figure reports that its end-of-line yield, basically the percentage of robots that pass final inspection, is now over 80%. Their battery production is even more efficient, boasting a 99.3% first-pass yield.

Before any Figure 03 robot is cleared for duty, it must undergo a “burn-in” session that sounds more like a gym workout than a factory test. The robots undergo over 80 functional tests, including thousands of squats, shoulder presses, and jogging sessions to catch any early mechanical hiccups.

Every new unit is a data-gathering machine for Figure. The growing fleet is feeding massive amounts of information back into Helix, the company’s AI model. By deploying more robots in the field, the company can identify and fix edge-case failures. This feedback loop allows the engineering team to push out over-the-air updates to the entire fleet, making every robot smarter simultaneously.

New capability: Perception-conditioned whole-body control

Alongside production scaling, Figure has also announced an upgrade to its control system known as Helix System 0 (S0), a humanoid AI model focused on full-body movement.

Previously, S0 relied mainly on proprioception, meaning the robot understood its own joint positions and body motion, but had limited awareness of its surroundings. That made navigation in complex environments like stairs or uneven terrain difficult without manual switching or intervention.

The new update introduces camera-based perception. RGB images from onboard cameras are processed into a 3D representation of the environment and combined with the robot’s internal body state. This allows the robot to both “feel” and “see” its environment simultaneously.

Figure says the system is trained using reinforcement learning in simulation across thousands of different terrains. Importantly, the behaviors learned in simulation transfer directly to real-world robots without additional fine-tuning or calibration.

In practice, this means robots can navigate stairs and uneven surfaces with human-like motion and stability, even under different lighting conditions, without operator input. The company describes this as a step toward more general-purpose robotic movement, where perception and control are tightly integrated into a single learning system.

Also read: Amazon’s Fauna Robotics acquisition shows how major tech companies are pushing humanoid robots closer to home use and AI-powered assistance.