A world model covering general manipulation and mobility, one of three R&D directions alongside Agentic OS and the humanoid body
Wearable collection device, ~470 g, trajectory precision <1 cm, multi-device sync latency <1 ms, over 30,000 tactile detection points
Skill collection terminal supporting standardized collection of complex skills
Data platform covering the full workflow of data management, annotation, quality inspection and release
Product system of 1 core device + multi-scenario skill collection solutions (per original notes)
High-precision micro-manipulation dexterous hand; demonstrated the 'threading a needle' feat in 2024 — an industry benchmark product; also available as L20 Lite and an industrial L20 version
High-end linkage-route model with high payload and precision
Entry-level dexterous hand; lightweight and extremely cost-effective
High-DoF dexterous hands; flagship models covering the tendon/direct-drive routes
Embodied AI foundation model that decomposes natural-language instructions into task chains and composes atomic skills to generate manipulation sequences
Teleoperation system; a teleoperated data collection solution building a data collection ecosystem
Skill library; billed as the world's largest dexterous hand manipulation dataset, with 500+ types of human skill data accumulated across industry, research and daily life
Dexterous manipulation robot; mass production has started, focused on dexterous manipulation needs in logistics scenarios
Member of the humanoid product line, for industrial collaboration and research education scenarios
Humanoid robot products covering home companionship and other scenarios
Integrated innovation solutions for manufacturing and service scenarios, with self-developed bodies, servo drives, and sensors
Hotel-scenario embodied service robot handling item delivery, public area patrol, room service, and basic cleaning; validated in a real hotel POC
Lightweight world model improving long-horizon task planning, cross-embodiment zero-shot adaptation, non-standard scenario manipulation generalization, and instruction following
Memory augmentation for complex long-horizon task planning and non-standard scenario generalization