UNIUBI's Cyvet Quadruped Shows New Leaps at WRC, 30 Days After Debut

During August 19–23, UNIUBI AI's fully in-house developed quadruped robot, the Cyvet, appeared at the 2026 World Robot Conference (WRC). This came just 30 days after its debut at the World Artificial Intelligence Conference (WAIC).

The robot has now unlocked two new athletic skills: a 1-meter Fosbury flop and a nearly 2-meter standing long jump. Combined with the world's first 720° consecutive backflip group performance, the Cyvet demonstrated rapid iteration and a commitment to expanding the capabilities of quadruped robots.

At WAIC 2026, the Cyvet drew attention with the world's first 720° consecutive backflip and a creative group dance. Just 30 days later, at WRC, it upgraded again—adding a 1-meter Fosbury flop and a nearly 2-meter standing long jump, expanding from flips to leaps.

This new jumping ability transforms the question from "can it do it?" to "can it be used?" The robot can now easily step onto platforms under power pylons, bypass rough gravel paths in chemical parks, and clear drainage ditches in factory areas without human intervention. With its 1-meter high jump and 2-meter long jump, these obstacles become traversable sections. UNIUBI's self-developed autonomous navigation lets it act as a confident inspector in substations, while the high-dynamic capabilities from flips and jumps serve as a powerful technical statement on the WRC stage.

Behind the rapid 30-day evolution is the competitiveness of full in-house development: from joint motors, motion control algorithms, visual perception, to multimodal interaction large models and inspection SaaS platforms—all independently developed without third-party licenses. The self-developed heterogeneous dual-brain, combined with high-performance joint modules, enables independent perception, reasoning, and decision-making even offline. The 15 kg chassis can carry a 20 kg load, and the ultra-low-power motion control model supports 5.5 hours of battery life. Flips, jumps, and loaded walking all rely on a single unified motion control model without switching policy networks.

When underlying capabilities are self-controllable, the pace of technical evolution is fully in its own hands. From WAIC to WRC, it iterated capabilities in 30 days; from algorithms to motors, full-stack in-house development builds a technical moat. The Cyvet is following the path of embodied intelligence, gradually turning each technical breakthrough into real value in industrial settings.

It took only 8 months for the Cyvet to go from design sketch to engineering mass production. This relies on UNIUBI's well-established product development, manufacturing, and delivery capabilities, forming a moat in the robotics track.

With 12 years of deep cultivation, starting from facial recognition and rooted in real-world scenarios such as construction sites, factories, buildings, communities, AI parking lots, border security, and autonomous driving, UNIUBI has accumulated technology, engineering, and application experience. Its products cover over 90 countries and regions, and it has honed a mature team.

After 12 years, UNIUBI possesses a rare self-developed embodied "brain" in the industry and launched its first heavy-duty quadruped robot. Targeting real roads and complex spaces, it effectively completes an end-to-end closed loop from perception, localization, path decision, to safe execution, moving from machine eyes to intelligent bodies. With 127 national authorized invention patents and an engineering implementation system repeatedly validated in projects, it gives the Cyvet the confidence to go from design to mass production in 8 months.

Now, the Cyvet is moving from exhibition to production lines, accelerating mass production and delivery. It will first be deployed in real scenarios such as community inspection, ushering in the liberation of human productivity in the real world.