Delta Intelligence demonstrates nine linked household tasks

What happened

A Beijing government science portal, citing the company's release, says the robot tidied a bed, moved toys and clothing, loaded a cup into a dishwasher, inspected fruit in a refrigerator and used a pedal bin before sitting down. Delta says the model combines visual, language and body-state inputs with whole-body control and was trained through a real-to-simulation-to-real process using physical interaction, human corrections and reinforcement learning.

Why it matters

The demonstration links navigation, manipulation and task switching on physical hardware rather than showing an isolated movement. It remains a company-selected technical demo: no independent benchmark, repeated-trial success rate, intervention count, safety evaluation or home deployment was reported.

Source checked October 31, 2026. Status describes the evidence in the linked source at the time of review.