KUAVO-MY vs Green: Key Differences

Green is a 180 cm, 75 kg humanoid with anthropomorphic hands: 12 of its 41 degrees of freedom are in the fingers and 4 in the wrists. KUAVO-MY is smaller and lighter at 166 cm and 55 kg, with 22 degrees of freedom on the standard version (28 with head and dexterous hands), and Leju focuses on whole-body motion control for research use.

Green walks faster at 0.9 m/s against KUAVO-MY's 0.4 m/s, although KUAVO-MY can also run at up to 5 km/h. Green navigates on its own in unfamiliar indoor spaces and steers around obstacles using its cameras and LiDARs.

Green carries eight RGB cameras and two LiDARs, plus inertial sensors for balance and force-torque sensors in its fingertips. KUAVO-MY combines a Livox Mid-360 3D LiDAR and an Orbbec Gemini 335L depth camera with a six-microphone array that locates sounds in 360°. Green's fingertip force sensing is the clearer differentiator for handling small or fragile objects.

KUAVO-MY uses Leju's KaihongOS embodied AI algorithms and supports teleoperation and learning-based control. Green runs Sber's GigaChat for conversation and the Green-VLA model (about 1.3 billion parameters, open weights) for manipulation, and it can detect and correct its own mistakes. Green is the stronger pick if natural voice interaction matters.

Sber says Green runs on a 48 V battery but has not published its capacity, runtime or charging time. KUAVO-MY's 60 V, 6 Ah pack gives about an hour of walking and recharges in 1.5 hours or less, so it suits demos and research sessions more than full shifts.

KUAVO-MY is aimed at robotics research, industrial testing and service tasks such as customer guidance and tote carrying. Green is positioned for paid work in shops (picking items and stocking shelves), factories (assembling and packing), labs (handling test tubes) and room service, with Sber planning pilot projects with businesses.

KUAVO-MY's ecosystem is KaihongOS, which makes it a fit for teams developing on OpenHarmony. Green's software is built on Sber's GigaChat and the open-weight Green-VLA model, and Sber's Green Challenge lets outside teams train controllers in a simulator of the robot. Sber has not announced a public SDK.

Sber has not published safety certifications for Green; it relies on balance sensing, obstacle avoidance and self-correction of errors. Neither maker lists formal certifications in our records, so organisations planning to run either robot near people should request safety documentation directly.