Figure 02 vs AgiBot A2: Price, Payload & AI Showdown (2026)

Figure 02 from Figure AI (USA) and the AgiBot A2 from AgiBot (China) represent two of the most ambitious humanoid platforms targeting industrial deployment this year. One arrives with a high‑profile BMW pilot and a multimodal AI brain; the other redefines the payload‑to‑weight ratio with 49 degrees of freedom. On paper, the AgiBot A2 offers substantially more lifting power, a lighter frame, and a $30,000 price advantage — yet the Figure 02 counters with a proprietary Helix AI stack, direct manufacturing trials, and a US‑based ecosystem. This article compares the two robots across every dimension a serious buyer needs, from deep specs and software readiness to real‑world deployment and total cost of ownership.

Quick Verdict: Who Wins?

Figure 02 vs AgiBot A2: Specs Compared

The numbers tell a dramatic story. The AgiBot A2 leaps ahead in payload, degrees of freedom, and top speed, while both platforms share similar height and 5‑hour runtimes. This table breaks down the known specifications as of mid‑2026.

The payload gap is the decisive hardware differentiator: the AgiBot A2 lifts four times more than the Figure 02, while weighing 15 kg less. It also boasts 21 extra degrees of freedom — a massive articulation advantage that suggests finer whole‑body control. On the software side, however, almost everything beyond the Helix AI stack remains opaque for both, forcing buyers to bet on the manufacturer’s roadmap.

Figure 02: AI‑Powered Industrial Humanoid from the US

Figure 02 is the second‑generation humanoid from Figure AI (Sunnyvale, California). Unveiled in 2024, it entered pilot operations at BMW’s Spartanburg plant later that year, where it performed sheet‑metal insertion and logistics tasks. The robot stands 1.68 m tall and weighs 70 kg, with 28 degrees of freedom, a 20 kg payload, and a 5‑hour battery. Its headline feature is the Helix multimodal AI model, which combines vision, language, and action reasoning to let the robot understand verbal instructions and adapt to unstructured manufacturing environments. Pricing is estimated at $130,000 for pilot‑program units; no public retail channel exists yet.

Figure 02’s value proposition is cognitive: it is designed to reason about tasks, handle exceptions, and communicate with human workers through natural language. The BMW deployment — though limited in scale — provides a real‑world validation of its AI‑driven approach. For organisations that need a humanoid that can learn new workflows on the fly and integrate into a Western supply chain, Figure 02 is the clear front‑runner.

AgiBot A2: The Chinese Payload Champion

AgiBot A2 is a bipedal humanoid from AgiBot (Shanghai, China), launched in 2025. It matches the Figure 02’s 5‑hour battery life and similar stature (1.69 m), yet weighs only 55 kg — 15 kg less — while carrying an astonishing 80 kg payload. With 49 degrees of freedom, it achieves a top walking speed of 3.3 m/s. At $100,000 pre‑order, it costs 23% less than the Figure 02. AgiBot has positioned the A2 as a multipurpose industrial workhorse, targeting logistics, manufacturing, and heavy‑lifting tasks where brute force matters more than high‑level reasoning.

The A2’s extraordinary payload‑to‑weight ratio (145% of body weight) and dense articulation suggest advanced actuator design, though detailed sensor and compute specs remain unpublished. The robot is available for pre‑order through the manufacturer and select distributors, with deliveries expected in 2025–2026. Its main limitation is the relative opacity of its software ecosystem and the unfamiliarity of its AI stack to Western integrators.

Price & Configurations

At the headline level, the AgiBot A2’s $100,000 price tag undercuts the Figure 02’s $130,000 by a significant margin. Both figures represent base‑unit costs for early‑adopter programmes; real deployment costs will rise with software licenses, end‑effectors, and integration services.

• Figure 02: $130,000 includes the robot hardware and access to the Helix AI platform for BMW‑style pilot tasks. Enterprise support, custom end‑effectors, and fleet management tools are negotiated separately. No lease‑or‑RaaS model has been announced, though Figure AI has indicated interest in a service‑based offering for future iterations.
• AgiBot A2: $100,000 secures a standard unit with a basic software stack and warranty. Custom grippers, advanced autonomy packages, and integration support are extra and priced on a per‑deal basis. AgiBot has hinted at volume discounts for fleet orders, but no public RaaS program exists yet.

For a three‑year total‑cost‑of‑ownership estimate (hardware + annual software/support at ~10‑15% of MSRP + spare parts), the AgiBot A2 likely lands around $130,000–$140,000, while the Figure 02 approaches $170,000–$180,000. That $30,000‑40,000 gap is meaningful for budget‑constrained operations, but the Figure 02’s pilot‑proven AI stack may shorten the time‑to‑useful‑work, partially offsetting the higher hardware cost.

Winner: AgiBot A2 delivers substantially more hardware capability per dollar, but Figure 02’s AI maturity could reduce integration expense. For pure capital‑efficiency, the AgiBot A2 wins; for total‑cost‑sensitive AI‑dependent workflows, the Figure 02 may still be the better bet.

Performance & Specs

Payload

The AgiBot A2’s 80 kg payload eclipses the Figure 02’s 20 kg by a factor of four. That transforms the robot from a tool‑handler into a genuine material mover. An A2 can lift a 55 gallon drum, a full pallet of small parts, or heavy automotive sub‑assemblies; the Figure 02 is limited to lighter tooling and components.

Winner: AgiBot A2.

Speed

The A2’s 3.3 m/s top speed (≈7.4 mph) versus the 02’s 3.0 m/s gives it a 10% edge, worthwhile for large‑footprint facilities. Both walk at speeds that outpace a human’s brisk walking gait.

Winner: AgiBot A2.

Battery & Runtime

Both robots achieve 5‑hour continuous operation, a common baseline for a half‑shift of work. Neither manufacturer has released swappable‑battery or fast‑charge details, so multi‑shift deployments will require a second unit or a still‑to‑be‑announced rapid‑charge solution.

Winner: Tie — identical runtime.

Degrees of Freedom (DOF)

With 49 DOF to Figure 02’s 28, the AgiBot A2 possesses significantly more articulation. The extra joints are likely concentrated in the arms, hands, and torso, enabling finer manipulation, broader reach envelopes, and more human‑like whole‑body motions. In practice, this means the A2 can do the mechanical work of two or three simpler robots.

Winner: AgiBot A2 — the DOF advantage is transformative for dexterous tasks.

Software & SDK Ecosystem

Figure 02’s defining software asset is the Helix multimodal AI, which processes natural language commands, visual scene understanding, and action planning to execute tasks learned from demonstration or verbal instruction. Access to Helix is currently limited to pilot partners; no public SDK exists. AgiBot A2’s software stack is a black box — early press materials mention a “GO‑1” AI model, but no detailed APIs, simulation tools, or developer documentation have been released. Neither robot offers ROS 2 integration out of the box.

Winner: Figure 02 by a wide margin. The Helix model is at least partially characterised and demonstrated. AgiBot’s silence makes its software readiness a gamble.

Autonomy & Sensors

Figure 02 relies on a head‑mounted RGB‑D camera cluster, solid‑state LiDAR, and the Helix visual‑language model for perception and navigation. It can map its environment, recognise objects, and adapt to changes without external markers. AgiBot A2 integrates binocular vision and force/torque sensors throughout its 49‑DOF body, but its full sensor suite and underlying SLAM/navigation stack are unpublished. Both robots lack IP ratings, limiting them to controlled indoor environments for now.

Winner: Figure 02 — its sensor‑to‑AI pipeline is better documented and has been seen operating in a real factory.

Build Quality & Durability

Both humanoids are built from lightweight alloys and composites. The AgiBot A2’s 55 kg frame, despite its 80 kg payload, suggests a highly engineered actuator and structural design that likely uses harmonic drive gearsets or high‑torque brushless motors. Figure 02’s 70 kg chassis, while heavier, may include more protective casing and a larger cooling system for the onboard NVIDIA compute cluster. Neither robot has published IP ratings or operating temperature ranges, so durability claims are untestable.

Winner: Tie, pending real‑world reliability data.

Price & Total Cost of Ownership

Factoring in software, support, and spares over three years, the AgiBot A2’s TCO advantage grows to roughly $30,000–$40,000 versus the Figure 02. However, if a company’s primary need is AI‑enabled flexible automation, the Figure 02’s tighter integration of Helix may reduce the cost of programming and re‑tasking. For heavy lifting, the A2’s fourfold payload advantage directly means fewer robots are needed to move a given volume of material, magnifying the per‑unit savings.

Winner: AgiBot A2 for capital‑constrained high‑payload deployments; Figure 02 when AI‑driven task flexibility offsets hardware costs.

Which Should You Buy? Figure 02 vs AgiBot A2 by Use Case

• Choose Figure 02 if…
• You need a humanoid that can understand voice instructions, adapt to changing tasks, and operate semi‑autonomously in a Western manufacturing environment.
• You are already working with Figure AI or BMW and want a seamless expansion of a Helix‑powered fleet.
• Your payloads stay under 20 kg and you value advanced AI reasoning over brute strength.

• Choose AgiBot A2 if…
• Your use case involves heavy material handling — palletising, machine tending with heavy workpieces, or cargo transfer.
• You need maximum articulation (49 DOF) for complex assembly or maintenance tasks.
• You are comfortable with a pre‑order timeline and are willing to develop a custom software layer on top of AgiBot’s hardware until its SDK matures.

• Choose neither if…
• You need an off‑the‑shelf, ROS‑native platform with a thriving developer community; both robots are still closed‑ecosystem.
• Your environment requires an IP65 or higher rating — neither has published weather‑proofing or dust resistance.

Real‑World Deployments & Track Record

The Figure 02 has logged pilot hours at BMW’s Spartanburg plant, performing sheet‑metal insertion and autonomous material movement. BMW and Figure AI have jointly shown the robot operating alongside human workers in a mixed‑model assembly line (Figure AI press release, January 2025). No total‑units‑shipped or operating‑hour figures have been disclosed, but the partnership is an active, named deployment.

AgiBot A2’s deployment history is far less transparent. The company has published videos of the A2 lifting heavy pallets and moving through warehouses, but no named industrial customers or quantified operating data have been made public as of June 2026. AgiBot’s previous generation A1 was demonstrated in Chinese smart factories, but scaling data remains proprietary.

Winner: Figure 02 — a tangible, named, ongoing industrial pilot carries far more weight than unsupervised promotional footage.

SDK, Software & Developer Ecosystem

Figure AI’s Helix model is accessed through a Python API, currently limited to partners. Third‑party developers cannot write applications for the Figure 02 without a direct relationship with the company. AgiBot has not released any public SDK, ROS 2 drivers, or simulation models for the A2. Both manufacturers have indicated that a broader developer programme may follow commercialisation, but for now, the ecosystem is a one‑player affair.

Winner: Figure 02, marginally, because its API exists and has documentation (even if closed); AgiBot offers zero developer‑facing material at this stage.

Support, Warranty & Availability

Figure 02 is available for pre‑order with a bespoke enterprise support agreement that covers hardware service, software updates, and the Helix API. The pilot‑phase support model has been hands‑on, with Figure AI engineers embedded at BMW’s site. AgiBot offers a standard manufacturer warranty for the A2, with terms and spare‑part lead times undisclosed. Both robots are pre‑order only, meaning delivery timelines are aspirational; the Figure 02 has the slight advantage of being in active pilots now, whereas the A2’s “2025–2026” delivery window still carries uncertainty.

Winner: Tie on concrete data; Figure 02’s active pilot may provide more confidence for near‑term availability.