Unitree G1 vs Figure 02: Price & Capability (2026)

The $13,500 Unitree G1 has already shipped to labs worldwide, making humanoid robotics accessible to researchers. The $130,000 Figure 02 is still in pre‑order and pilot phases, but it brings a massive payload, advanced multimodal AI, and an industrial design that puts it in a different league. This article breaks down every dimension—specs, price, software, real‑world deployments, and total cost of ownership—so you can decide whether the affordable budget bot or the AI‑powered industrial workhorse matches your needs.

Quick Verdict: Unitree G1 vs Figure 02

Unitree G1 vs Figure 02: Specs Compared

The table below captures the core hardware facts. All figures are drawn from manufacturer specifications and verified in mid‑2026.

What jumps out: The Figure 02 more than quadruples the G1’s payload, runs over twice as long, and reaches a higher top speed—yet it costs roughly ten times as much and cannot be purchased outright.

Unitree G1 Overview: Affordable Research Humanoid

Launched in 2024 by Unitree Robotics in Hangzhou, China, the Unitree G1 is the company’s first mass‑market humanoid. It is available in three configurations: the entry‑level Base model ($13,500), the EDU version with an NVIDIA Jetson Orin module and enhanced sensors (~$43,900), and the Ultimate variant with the full 43‑DOF dexterous‑hand option (~$73,900). With a compact 127–132 cm height and 35 kg weight, the G1 is designed for table‑top manipulation, locomotion research, and light service tasks. Over 1,000 units are reported to have shipped to universities, labs, and developer communities worldwide, making it one of the most accessible full‑body humanoids on the market.

Figure 02 Overview: AI-Powered Industrial Workhorse

Figure 02 is the second‑generation humanoid from Figure AI, a Sunnyvale, California startup backed by $675 million in funding. Unveiled in August 2024, the Figure 02 stands 1.68 m tall, weighs 70 kg, and packs 20 kg of payload per arm—enough to handle automotive parts and heavy tooling. It is powered by Helix, a proprietary multimodal vision‑language‑action model that lets the robot observe, understand, and execute tasks through natural language instructions. Figure 02 is not sold publicly; it is deployed through pilot programs, most notably at BMW’s Spartanburg manufacturing plant, where it has performed autonomous part‑handling tasks. Pricing is estimated at $130,000 per unit when commercial orders open, likely in late 2026 or 2027.

Price & Configurations: $13,500 vs $130,000

The raw sticker gap is over 9x, but the configuration differences amplify it further. Unitree offers a clear stair‑step model: $13,500 for the Base (23 DOF, basic hands), $43,900 for the EDU (adds Intel RealSense D435i, NVIDIA Jetson Orin NX, and 3‑finger gripper), and $73,900 for the Ultimate with two 6‑DOF dexterous hands and a full‑suite sensor package. The Figure 02 comes in a single, fully‑loaded configuration that includes 28 DOF, five‑fingered hands, Helix AI, and a complete perception stack for the estimated $130,000. There is no lesser‑equipped “light” version.

For a research lab, a single G1 EDU costs less than a third of a Figure 02, and the Base model costs one‑tenth. Winner: Unitree G1 for raw affordability and configurability.

Performance & Specs Face-Off

Payload & Strength

The G1’s 3 kg arm payload limits it to light objects—think groceries, small tools, or lab pipettes. The Figure 02’s 20 kg rating lets it lift a car‑door assembly, heavy totes, or industrial tooling. Winner: Figure 02.

Speed & Mobility

Figure 02 reaches 10.8 km/h on paper, outpacing the G1’s 7.2 km/h, though top‑speed demonstrations in factory settings are rare. Both walk bipedally and navigate uneven terrain, but the heavier Figure 02 benefits from a more robust lower body. Winner: Figure 02 for raw speed potential.

Battery Life & Endurance

A 5‑hour battery lets the Figure 02 work an entire shift segment without recharging. The G1’s 2‑hour runtime limits continuous operation and increases downtime in a production environment. Winner: Figure 02.

Degrees of Freedom & Dexterity

The G1 can be configured up to 43 DOF (including dexterous hands) on the Ultimate tier, but the Base model has only 23. Figure 02’s fixed 28 DOF includes five‑fingered hands that can manipulate complex parts. For fine manipulation, the Figure 02’s dedicated hands and Helix’s control policy give it an edge, though the G1’s optional 6‑DOF hands on the Ultimate model level the field somewhat. Winner: Figure 02 (for real‑world dexterity out of the box).

Software & SDK Ecosystem

Unitree publishes a Python/C++ SDK and supports ROS 2 out of the box. The G1 runs on a Linux‑based system, and the EDU/Ultimate variants ship with the NVIDIA Jetson Orin, making it a natural fit for AI‑based research. Hundreds of community projects, simulation models (Isaac Sim, Mujoco), and open‑source repositories exist. In contrast, Figure AI has not released a public SDK or ROS 2 interface as of mid‑2026. The Helix model is proprietary, and developers can only interact through controlled pilot APIs. For anyone who needs to write custom autonomy or integrate with existing stacks, the G1 is the only open option. Winner: Unitree G1.

Autonomy & Sensors

The G1 relies on depth cameras, an IMU, and optional LiDAR for SLAM‑based navigation; it can be programmed with Unitree’s Autowalk and custom planners. The Figure 02 employs six RGB cameras, depth sensors, LiDAR, and an onboard Helix AI engine that merges visual perception with natural‑language commands. In BMW’s Spartanburg plant, Figure 02 automatically picks body‑shop parts from a rack and places them on a conveyor, adapting to positional variance—a level of adaptive autonomy that surpasses pre‑programmed pick‑and‑place. While the G1 can be made to do similar tasks with off‑board compute, the Figure 02’s integrated AI stack is purpose‑built for flexible automation without external coding. Winner: Figure 02 for out‑of‑the‑box autonomous task execution.

Build Quality & Durability

Unitree’s G1 is built primarily for indoor, climate‑controlled environments. Its chassis uses lightweight alloys and plastics; an IP rating is not published, but it is not designed for dust or water exposure. The Figure 02, with a 70 kg metal‑framed body and industrial connectors, is built to survive factory floors. While an official IP rating is also undisclosed, the robot’s BMW deployment implies it can handle the mild grime and temperature swings of a manufacturing environment. For long‑term, continuous duty cycles, the Figure 02’s construction is clearly more robust. Winner: Figure 02.

Price & Total Cost of Ownership

A three‑year TCO snapshot illustrates the gap beyond the hardware price:

Unitree provides a 1‑year warranty and sells spare parts openly; replacement arms or actuators cost a few hundred dollars. Figure AI’s support model is embedded in its pilot agreements, with service and software updates included, but spare parts are not yet available to third parties. For organizations that need a fleet, the G1’s hardware‑only cost means you could buy three EDU units for less than one Figure 02 pilot slot—and have full control over maintenance. Winner: Unitree G1 for predictable, transparent TCO and spares availability.

Which Should You Buy? Unitree G1 vs Figure 02 by Use Case

Choose the Unitree G1 if you… - Are a university lab, maker space, or STEM educator that needs an affordable, hackable humanoid platform. - Require ROS 2 and an open SDK for locomotion or manipulation research. - Have a budget under $50,000 and want a robot you can order and receive within weeks. - Plan to integrate custom AI models on the Jetson Orin.

Choose the Figure 02 if you… - Operate a manufacturing line or industrial pilot where payload, runtime, and ruggedness are critical. - Need adaptive automation that can be guided by natural‑language instructions rather than coding. - Have a six‑figure budget and can wait for a dedicated pilot program. - Are investing in the Helix ecosystem and expect long‑term scalability across a fleet.

Choose neither if… - You need a proven, production‑ready robot for 24/7 manufacturing with established safety certifications (look at collaborative robots like the UR10e instead). - You require outdoor, all‑weather operation—neither robot is built for that.

Real-World Deployments & Track Record

Unitree G1: Since its launch in mid‑2024, Unitree has shipped over 1,000 G1 units to more than 40 countries. Key installations include Tsinghua University’s bipedal locomotion lab, Korea’s KAIST for human‑robot interaction studies, and dozens of independent developers and YouTube creators. There are no large‑scale industrial deployments; the G1 remains primarily a research and education platform.

Figure 02: After its 2024 unveiling, Figure AI placed the first Figure 02 units in BMW’s Spartanburg, South Carolina plant. In a trial running from early 2025, the robot performed autonomous “body‑shop logistics”—picking stampings from a stillage and placing them on a conveyor at a rate of several hundred parts per day. Figure has publicly stated the robot logged over 1,000 operating hours across multiple units without a significant failure (as reported by BMW in a joint press release, March 2025). No other public deployments have been confirmed; the robot is still in pilot phase.

Both platforms are unproven in full‑scale, multisite production, but the Figure 02’s factory data gives it a slight real‑world edge for industrial relevance. Winner: Figure 02 for demonstrated industrial operation hours.

SDK, Software & Developer Ecosystem (Deep Dive)

Beyond the basic ROS 2 and language support, the development experience diverges sharply:

• Unitree G1: GitHub hosts dozens of open‑source repos for G1 control, with Isaac Sim and MuJoCo models actively maintained. Unitree’s own repository provides a Teleoperation SDK, full‑body motion retargeting, and an AI‑training pipeline that leverages the Jetson Orin. A growing Slack and Discord community provides peer support. Academic papers using the G1 for whole‑body control have appeared at ICRA and IROS 2025.

• Figure 02: No public SDK, no simulation models, no developer forum. Figure AI’s Helix model is accessible only to pilot partners via a gated API. Third‑party developers cannot write custom controllers, integrate new sensors, or modify the autonomy stack. While Figure’s internal capabilities are sophisticated, the lack of openness locks out the broader robotics research community.

For any organization that plans to customize the robot’s behavior, the G1 is the only viable choice. Winner: Unitree G1.

Support, Warranty & Availability

Unitree G1: Available for immediate purchase through Unitree’s website and authorized distributors like Robot Overflow. Lead time is typically 2–4 weeks. Standard warranty is 1 year, with extended service plans available. Replacement parts (arms, actuators, battery packs) can be ordered directly. Enterprise support with a service‑level agreement is offered for bulk orders of the EDU/Ultimate configurations.

Figure 02: Not yet available for public purchase. Interested enterprises must join the pilot waitlist. Figure AI provides full‑service support and maintenance as part of the pilot agreement, but warranties, spare‑parts pricing, and standard SLAs have not been publicly released. Expect lead times of 6–12 months once commercial orders open. Winner: Unitree G1 for immediate availability and transparent support terms.