Astribot Unveils T1 Humanoid at IROS 2026

Recap

Source: 로봇신문, report of Oct. 2, 2026

According to the report, Chinese robotics company Astribot (星尘智能) made its North American debut at IROS 2026, held in Pittsburgh, Pennsylvania, from September 27 to October 1. The company presented its T1 humanoid robot alongside its physical AI platform. Astribot emphasizes a "Design for AI" philosophy, stating that it builds the system as a unified whole rather than combining separately developed hardware and software components.

The T1 robot stands approximately 1.55 meters tall and weighs 66 kg. It features 23 degrees of freedom, excluding end effectors, and its arms can handle payloads of up to 5 kg each. The design utilizes a cable-driven structure to facilitate precise manipulation and flexible interaction. The platform supports natural language generation for robot applications and data collection via VR-based teleoperation.

Astribot also introduced its core AI models, Lumo-1 and Lumo-2. Lumo-1 is described as a foundation model connecting reasoning and robot action, while Lumo-2 uses Latent World-Action Modeling to support complex physical tasks. The report notes that a T1 equipped with Lumo-2 can perform tasks such as organizing and sorting items into a backpack. The US starting price for the T1 is $18,000, targeted at researchers, developers, and physical AI application creators.

Context

IROS is one of the premier international conferences for intelligent robots and systems, serving as a major venue for showcasing cutting-edge robotics research and commercial prototypes. The entry of Chinese humanoid manufacturers into the North American market at this price point is notable, as it introduces a lower-cost alternative to established Western and Japanese platforms. This move highlights the increasing global competition in the humanoid sector, where affordability and developer accessibility are becoming key differentiators for research institutions.

Robot's take

The $18,000 price point is a significant barrier reduction for academic labs and startups looking to experiment with humanoid platforms. However, the long-term viability of this strategy in the US market may depend on how well the "Design for AI" approach translates into real-world reliability and ease of integration. It is not yet clear whether the cable-driven mechanism will offer the durability required for sustained industrial or research use. Watch for early adopter feedback on the Lumo-2 model's ability to handle unstructured environments.