EEmbodied AI Hub
IndustryDeploymentEditor’s pick

AGIBOT Fields A3 Ultra Humanoid and G2 Max Task Platform at WAIC to Push Repeatable Industrial Workflows

AGIBOT launched four embodied AI products at WAIC 2026, headlined by the A3 Ultra full-size humanoid and G2 Max heavy-payload task robot, to shift industry focus from single demos to manufacturable, integrable systems. The Shanghai company, which has now built 15,000 units since its 2023 founding, pairs its Three Intelligences in One architecture with NVIDIA Thor processors to target commercial service and industrial material-handling deployments.

OmniHand 3 Ultra-Mrobotics-businessNVIDIA ThorPeng ZhihuideploymentA3 UltraAGIBOTG2 Max

Source: Robotics Business Review · July 20, 2026

Share this article so more people can see it

AGIBOT used its 200-square-meter booth at WAIC 2026 to present four new platforms aimed at moving embodied AI from laboratory showcases into sustained commercial and industrial operations. The Shanghai company positioned the releases as evidence that the sector must now measure progress by repeatable delivery and integration rather than isolated demonstrations.

Co-founder, president and CTO Peng Zhihui stated that the industry should judge embodied AI by whether robots can be manufactured, delivered and integrated reliably into real operating environments. He highlighted the A3 Ultra, G2 Max and OmniHand 3 Ultra-M as addressing the humanoid platform, industrial task execution and dexterous manipulation layers respectively.

Founded in 2023 as Zhiyuan New Technology Co., AGIBOT has already reached a production milestone of 15,000 robots. Its Three Intelligences in One architecture is designed to unify locomotion, interaction and manipulation within a single control stack, supporting both humanoid and non-humanoid form factors.

The A3 Ultra stands 1.74 m tall, weighs 60 kg and carries 51 active degrees of freedom with a 5 kg payload per arm. It is rated for up to eight hours of operation and incorporates 3D lidar, RGB-D and fisheye cameras plus binocular vision, backed by GPS, RTK and UWB positioning for multi-environment navigation.

Powered by NVIDIA Thor for high-level embodied AI processing, the A3 Ultra supports direct charging, battery swap and autonomous docking so it can remain available for extended service shifts rather than single appearances. AGIBOT presented the platform performing public-facing guidance tasks alongside dance demonstrations.

The G2 Max is the company’s first heavy-payload, force-controlled embodied task robot, built for material handling, palletizing and similar industrial operations. It combines higher load capacity with force-controlled arms, adjustable working height and omnidirectional mobility to fit existing factory layouts.

Complementing the larger platforms, the OmniHand 3 Ultra-M supplies direct-drive dexterous manipulation while the X2 Edu variant targets educational users. AGIBOT displayed all three commercial products together with manipulation foundation models and live industrial deployment footage inside its exhibition space.

The company’s emphasis on production volume and factory deployments contrasts with many peers still focused on prototype capability. By releasing hardware explicitly sized for commercial service and heavy industrial tasks, AGIBOT is signaling that capital is now being allocated to scale manufacturing and integration rather than further capability research alone.

For builders, the choice of NVIDIA Thor and the inclusion of multiple charging modes indicate a hardware bet on long-duration autonomy and standardized compute. The rapid production ramp to 15,000 units also suggests that closed-loop data collection from deployed fleets is becoming a near-term priority for model improvement.

Source: Robotics Business Review.

Related resources on this hub

Jump to projects, models, or datasets mentioned or closely related.

Discussion

Tell us what you think — comments make stories more useful for builders and founders.

Tell us what you think!

AGIBOT Fields A3 Ultra Humanoid and G2 Max Task Platform at WAIC to Push Repeatable Industrial Workflows

Have an account? Log in to use your display name and avatar.

Email is optional and never shown on the page.

More insights

IndustryDeploymentRobotics Business Review

Beyond Vision and Force: Why Friction is the Missing Link in Robot World Models

World models are hitting a physical wall in contact-rich manipulation because they lack precise friction data at the contact interface. The emergence of the VμA model class and advanced tactile hardware like Contactile are making the coefficient of static friction a first-class input, shifting the hardware-software co-design paradigm for embodied AI deployment.

Read
IndustryDeploymentRobotics Business Review

Burro Prioritizes Infrastructure-Free Reliability for Outdoor Agricultural Deployments

Burro shifted early from controlled demos to building robots that function in unpredictable outdoor agricultural settings without fixed infrastructure or reliable GPS. The company learned that user dependency forms rapidly once value appears, demanding zero tolerance for failures across extreme environmental variability. This approach targets the gap between short-horizon proofs of concept and sustained Tuesday-morning operations in rain, dust, and canopy conditions.

Read
IndustryDeployment量子位

Chaowei Power and Peking University Healthcare: A Pragmatic Path for Embodied AI in Hospitals

Chaowei Power and Peking University Healthcare partner to build a step-by-step embodied AI deployment path for hospitals, focusing on simulation, validation, and application. The collaboration uses Chaowei's world model engine and data collection headset to create high-fidelity training environments, addressing the lack of real medical training data. The approach prioritizes safety over speed, targeting tasks like tube sorting and drug delivery.

Read