The brief
- US debut at CES 2026 with GR-3, 165 cm humanoid with 55 DoF for eldercare, rehabilitation and companionship.
- Raised $109M in Series E funding (800M yuan) January 2025; additional $41.7M corporate investment May 2025.
- GR-3 features hybrid AI (fast reflexes + LLM reasoning), multimodal sensors (vision, audio, touch) for natural emotional interaction.
Technical approach
Care-first design
Humanoid optimized for eldercare and companionship
Unlike industrial humanoids, GR-3 prioritizes approachability, emotional expression and safety for vulnerable populations.
Full-Perception system
Integrated vision, audio and touch sensing
Structured-light camera module for facial recognition and eye contact, four-microphone array for voice localization, 31 pressure sensors for touch perception.
Hybrid AI control
Fast reflexes plus language model reasoning
Simple commands use fast rule-based control; richer conversations route through LLM for context-aware interaction.
Expressive movement
Multiple gaits and postures for emotional expression
Straight-leg walking, running, bouncy walks and varied postures enable robot to convey mood and engagement naturally.
