Meet GR-3: Fourier’s First Care-Centric Humanoid Robot

Fourier has unveiled GR-3, its first full-size “Care-bot,” marking a significant evolution in its GRx humanoid series. This new class of AI robots is specifically engineered for human-centric scenarios, guided by a philosophy that prioritizes “love, above all functionality.” GR-3 aims to redefine human-machine interaction, distinguishing itself with a soft-touch outer shell and a sophisticated multimodal emotional interaction system designed to foster “warm tech companionship.”

Moving beyond the traditional utility-focused robot, GR-3 embodies the “Care-bot” concept, a machine built not merely to perform tasks but to forge connections. Its primary roles are envisioned as a social companion, capable of engaging people in public settings, offering a comforting presence for children, or acting as a responsive companion for older adults. Looking ahead, Fourier anticipates GR-3 expanding into assistive care, with potential applications ranging from mobility support and health monitoring to physical rehabilitation in eldercare and clinical environments, aligning with the company’s vision of a robot that not only works but also cares.

In a clear departure from typical industrial aesthetics, GR-3 embraces a softer, more lifelike appearance. Standing 165 centimeters tall and weighing 71 kilograms, its design features warm, neutral tones, cushioned surfaces, and premium automotive upholstery, creating an approachable presence for everyday human interaction. The robot offers up to 55 degrees of freedom, integrating essential sensors and computing units within a compact, modular architecture. Expressive design elements, including balanced limb proportions, fluid movement, and an animated facial interface, are meticulously crafted to foster user comfort and trust. Powering its continuous operation, GR-3 is equipped with a hot-swappable battery system and intelligent power management, ensuring uninterrupted service in real-world environments.

At the heart of GR-3’s human-like interaction capabilities is Fourier’s proprietary Full-Perception Multimodal Interaction System. This advanced system seamlessly integrates vision, audio, and tactile feedback into a unified real-time emotional processing engine. An attention management system dynamically prioritizes sensory inputs and synchronizes multimodal outputs, enabling smooth, humanlike responses. For audio perception, a four-microphone array provides omnidirectional voice activation, echo cancellation, and precise sound source localization, allowing GR-3 to orient toward speakers with synchronized eye contact, enhancing natural vocal exchanges. Visual recognition is handled by an RGB and structured-light camera module, facilitating reliable facial recognition and dynamic tracking to keep users within GR-3’s expressive visual field. Tactile feedback is facilitated by 31 distributed pressure sensors, enabling real-time touch detection. Combined with animated micro-expressions—such as blinking, eye tracking, and emotion-triggered gestures—GR-3 responds to touch with remarkably lifelike social cues. Its dual-path response architecture supports both rapid reflexive actions (“fast thinking”) via rule-based control and deep contextual understanding (“slow thinking”) powered by a large language model for complex, situational dialogue generation, allowing GR-3 to interpret emotion, recognize intent, and respond with genuine awareness.

Beyond its emotional intelligence, GR-3 features enhanced hardware and sophisticated locomotion, enabling more intuitive and lifelike interactions. Drawing on extensive training data and market feedback from previous models, GR-3’s whole-body structure has been redesigned for greater flexibility and a more compact form. It incorporates Fourier’s high-performance actuators and 12-degree-of-freedom dexterous hands, supporting a wide array of tasks from interaction and navigation to practical assistance. The robot demonstrates versatile motion capabilities, including a standard walking gait that supports bending and squatting, agile movement in complex environments, and seamless multi-task collaboration. It can shift between various modes, from a straight-leg gait and running to an energy-saving “fatigue mode” and a lively “bouncy walk” for dynamic scenarios. Fourier has also prioritized the developer experience, building GR-3 as an embodied intelligence platform that supports a wide range of algorithms and third-party applications. Its client-server architecture allows researchers, automation engineers, and AI developers to quickly tailor functions for real-world use, with future interaction APIs set to further lower the barrier to entry for personalized, emotionally responsive applications.

GR-3 represents a significant milestone in Fourier’s pursuit of human-robot integration. Every aspect of its design, from hardware architecture to software framework, has been shaped by the fundamental question of whether a robot can be genuinely accepted in human environments. As Fourier’s most advanced response to date, GR-3 extends beyond mere performance, built instead to support meaningful interaction across diverse real-world settings, from public services and academic research to clinical use and, eventually, personal spaces, adapting seamlessly to evolving needs and complex environments as it operates alongside people.