Astribot T1 humanoid hits North America at $18,000

Astribot's North American debut is built around a three-layer pitch: its Lumo family of AI models, an embodied operating system, and a cable-driven robotic body. Rather than treating AI as a software layer bolted onto an existing robot, the company says it designs intelligence, the operating system and the physical body as one integrated system.

The OS layer appears aimed squarely at developers. T1 offers SDK and API access to joint control, Cartesian motion and whole-body coordination, plus sensor data, while Astribot OS provides meta-packages and skill libraries. At the second Astribot OS Hackathon, which concluded September 21 in Beijing, 15 teams built and demonstrated more than ten functional applications on T1 in 36 hours, according to the company.

The company's upgrade loop may be its real differentiator, or its biggest open question. T1 does not learn autonomously during deployment; instead Astribot gathers multimodal data from operation and teleoperation, retrains models such as Lumo-2, and pushes skills out through over-the-air updates or on-site calibration. Whether customers see that loop translate into meaningfully better household task performance likely depends on how quickly that retraining cycle turns.