Beijing delivers 15 humanoid robots and a full-stack platform as human-robot half-marathon nears

Delivery, hardware and platform

Beijing Humanoid Robot Innovation Center (北京人形机器人创新中心) on March 19 formally handed over 15 “Tiangong” (天工) humanoid robots to representatives from top universities — Peking University (北京大学), Beihang University (北京航空航天大学, 北航), Beijing Institute of Technology (北京理工大学, 北理工), Huazhong University of Science and Technology (华中科技大学, 华中科大) and The Hong Kong University of Science and Technology (Guangzhou) (港科大（广州）) — and second‑stage development partners including NIO (蔚来) and China State Shipbuilding Corporation (中船). It has been reported that the center also opened its general embodied‑intelligence stack “Huisi Kaiwu” (慧思开物) to partners, offering an integrated hardware‑to‑software technology base.

Six of the units are described as “Embodied Tiangong 3.0” (具身天工 3.0), a general‑purpose humanoid platform advertised for millimetre‑level fine manipulation and high‑dynamic motion. Nine are “Embodied Tiangong Ultra” (具身天工 Ultra), engineered for extreme sport and high‑speed running to tackle complex, fully autonomous tasks. The package is being pitched as a development and testing bed for both academic teams and industrial integrators.

Marathon testbed and strategic significance

The delivery is timed ahead of the 2026 Beijing Yizhuang half‑marathon, scheduled to start on April 19, which will for the first time implement a “human‑machine co‑running” mode: about 12,000 human runners will compete on the same course as the robots. It has been reported that last year Tiangong won the world’s first humanoid half‑marathon in 2:40:42; this year the centre’s CTO Tang Jian says the goal is to reach professional champion levels. Crucially, the event will include more than 20 teams running in autonomous navigation mode, rather than being remotely guided as in prior competitions.

Beyond spectacle, the effort signals how China is using public competitions and university‑industry partnerships to accelerate embodied AI and autonomy. This push coincides with tighter Western export controls on advanced AI accelerators and chips, making domestic, full‑stack platforms — from actuators and sensors to perception and control software — strategically important for research, talent cultivation and commercialisation. Can human runners keep up? The answer will tell us as much about robotics engineering as about China’s industrial strategy.