Robotics World Just Shifted
As far as new Robotic Tech development announcements go, few have sparked as much buzz as Figure AI’s unveiling of their new Figure 03 yesterday. Billed as the “Model T of robots,” this third-generation humanoid marks a significant shift toward affordable, scalable domestic assistants, capable of tackling everyday chores with human-like finesse.
Standing at 5 feet 6 inches and weighing around 70 kilograms, the Figure 03 isn’t just another prototype, it’s designed for mass production, with ambitions to integrate the robot into its own assembly lines for self-sustaining manufacturing.
Powered by Figure’s proprietary Helix AI system—a vision-language-action model that processes real-time inputs for seamless task execution—the robot promises to redefine household automation. For robotics enthusiasts tracking the market’s maturation, the Figure 03 exemplifies how AI integration and hardware refinements are closing the gap between lab demos and living-room reality.
Figure 03 Sets a New Standard in Agility
At its core, the Figure 03 excels in practical utility. Demonstrations showcased it folding laundry with precise grips, loading dishwashers without mishaps, and navigating cluttered kitchens autonomously. Key upgrades include a revamped sensor suite: six RGB cameras deliver a 60% wider field of view per unit, enabling 360-degree environmental awareness without blind spots.
The hands, are a standout feature, boast ultra-sensitive tactile palms that detect pressures as low as a few grams—rivalling human fingertip sensitivity for delicate manipulations like handling eggs or threading needles. Wireless charging eliminates cumbersome cables, supporting up to five hours of continuous operation on a single top-up, while the softer, more ergonomic design reduces intimidation factors for home use.
Helix AI orchestrates these capabilities through end-to-end learning, allowing the robot to generalize from language prompts (“Clear the table after dinner”) to multi-step actions without rigid programming. Priced for scalability (though exact figures remain under wraps), Figure envisions fleets of these bots in homes by 2027, addressing labor shortages in aging populations and dual-income households.
Watch the Demonstration Video Click Here
To appreciate the Figure 03’s leap, it’s essential to trace Figure AI’s iterative progress. Founded in 2022 by ex-Tesla and Boston Dynamics veterans, the company has iterated rapidly from clunky prototypes to polished performers. The inaugural Figure 01, revealed in 2023, was a proof-of-concept: a 5’6″ bipedal frame with basic locomotion and manipulation, but heavily reliant on teleoperation for tasks. It boasted a 1.2 m/s walking speed and 20kg payload but suffered from limited onboard compute—relying on cloud processing for AI inference—and a runtime of just 2-3 hours on lithium-ion batteries. Autonomy was nascent; it could pick objects via scripted demos but struggled with unstructured environments.
Enter Figure 02, launched in August 2024, which addressed these pain points head-on. Tripling onboard computation with dual NVIDIA RTX GPUs, it enabled fully autonomous real-world interactions, including conversational dialogue powered by OpenAI integration. Battery life jumped 60% to over four hours via solid-state tech, and redesigned hands with embedded sensors improved dexterity for tasks like tool use. At 168cm tall and 70kg, it matched human proportions while hiding wires for a sleeker aesthetic. Early pilots at BMW factories demonstrated its industrial viability, sorting parts with 95% accuracy.
Advances in Models Compared
The Figure 03 builds directly on this foundation, emphasising consumer scalability over industrial ruggedness. Below is a comparative table highlighting key evolutions:
| Feature | Figure 01 (2023) | Figure 02 (2024) | Figure 03 (2025) |
|---|---|---|---|
| Height/Weight | 5’6″ / 70kg | 5’6″ / 70kg | 5’6″ / 70kg |
| Runtime | 2-3 hours (Li-ion) | 4+ hours (solid-state) | 5 hours (wireless charging) |
| Compute/AI | Cloud-dependent, basic NN | 3x inference (dual NVIDIA RTX) | Helix VLA model, 60% wider FOV |
| Hands/Sensors | Basic grippers, no tactile | Embedded palm sensors | Gram-level tactile, 6 RGB cams |
| Autonomy | Teleop-heavy | End-to-end neural, conversational | Real-time household generalization |
| Payload/Speed | 20kg / 1.2 m/s | 20kg / 1.2 m/s | 20kg / 1.2 m/s (optimized balance) |
| Target Use | Prototyping | Industrial (e.g., BMW) | Domestic (chores, self-production) |
This progression underscores Figure’s focus from validation in figure 01 to viability in figure 02 to scaling into volume potential in figure 03, with each iteration slashing production costs by 40-50% through modular design.
Competitive Market Driving Innovation
Yet, Figure isn’t alone in this humanoid arms race. 2025 has seen a flurry of launches, intensifying competition and driving innovations in AI dexterity and cost-efficiency. The latest Optimus Gen 3 iteration, teased at Tesla’s AI Day in June, features enhanced neural nets from autonomous driving tech, enabling fluid walking at 1.5 m/s and 25kg payloads.
Unlike Figure’s home focus, Optimus targets factories, with demos of assembling car parts at sub-$20,000 price points—undercutting Figure’s estimated $30,000 entry. However, it lags in tactile feedback, relying on vision alone for manipulations.
Other Market Challengers:
- Boston Dynamics’ all-electric Atlas, refreshed in early 2025, remains the agility benchmark. Ditching hydraulics for electric actuators, the 89kg bot performs acrobatics like backflips and 2 m/s sprints, with a 1-hour runtime and 11kg payload. Priced at $150,000+, it’s geared for R&D and logistics, boasting superior balance via reinforcement learning but limited AI for unstructured tasks compared to Figure’s Helix.
- Agility Robotics’ Digit, now in its 2025 “Digit 2.0” variant, emphasises warehouse deployment with over 100 units piloted at Amazon. At 1.2m tall and 40kg, it lifts 18kg boxes at 1.2 m/s, powered by a 4-hour battery and onboard ML for tote handling—95% uptime in trials. It’s more affordable ($100,000) but lacks full bipedal versatility, focusing on repetitive lifts over Figure’s dexterous chores.
- Sanctuary AI’s Phoenix, updated in March 2025, shines in human-like motion with 30+ degrees of freedom and Carbon AI for task learning. The 65kg, 5’7″ frame handles tools and conversations via haptic gloves simulation, with a 3-hour runtime and 15kg payload. At $50,000-$75,000, it’s a strong home/industrial hybrid, though its slower inference (cloud-heavy) trails Figure’s edge compute.
Future Market Projections Positive Despite Challenges
For robotics market watchers, these developments signal a maturing ecosystem: robotic humanoid shipments are projected to hit 100,000 by 2030, per IDTechEx, fuelled by falling actuator costs (down 70% since 2020) and AI commoditisation.
Figure’s consumer pivot differentiates it from Tesla’s scale and Boston’s flair, potentially capturing the $50B home robotics market share by 2035. There are however still major hurdles for these companies to overcome: Safety standards and protocols are still lacking, and ethical debates around job displacement are perhaps still going to be the largest challenge for the industry to address.
Integrations like Figure’s self-production model could accelerate adoption at a more rapid pace than Electric vehicle production did. As the next year’s developments unfold, the Figure 03 isn’t just a robot; it’s a harbinger of ubiquitous, intelligent companions reshaping labor and leisure.
