GR-2

Fourier GR-2 is a human-scale, developer-friendly humanoid built for fast iteration. It upgrades the earlier GR-1 with more joints, stronger actuators, and a cleaner modular design.

This makes it ideal for universities, robotics labs, and industrial R&D teams working on loco-manipulation and AI-driven behaviors.

Key Features and Benefits

• Human-scale platform: 175 cm tall and 65 kg, sized for real-world environments and human-level reach.
  
  
  

• 53-DOF body: enables more expressive locomotion and dexterous manipulation than earlier models.
  
  
  

• High-torque FSA 2.0 actuators: up to ~380 N·m peak torque for stronger, more precise joint control.
  
  
  

• Lab-ready runtime: ~2 hours per charge, enough for experiments, demos, and data collection.
  
  
  

• Broad software toolchain: supports NVIDIA Isaac Lab, ROS, and MuJoCo, plus VR teleop and lead-through teaching.
  
  
  

• Modular, serviceable design: integrated cabling and modular layout speed part swaps and customization.
  
  
  

• Verified listing: Humanoid Guide shows a verified entry, giving researchers confidence in baseline performance.
  
  
  

How It Helps

R&D teams can prototype perception-to-action loops, test skill libraries, or train policies in simulation and bring them onto the real robot. Because GR-2 supports ROS, MuJoCo, and NVIDIA Isaac Lab, labs can keep their current tools and shorten the path from virtual to physical experiments.

What’s Included / Platform Capabilities

The platform offers high-torque FSA 2.0 joints, integrated internal wiring, modular components, and multiple control modes (VR teleop, lead-through, direct commands).

This gives researchers flexibility to explore manual demonstrations, teleoperated data collection, or fully autonomous behaviors on the same machine.

Specifications

• Height: 175 cm
  
  
  

• Weight: 65 kg
  
  
  

• Overall DOF: 53
  
  
  

• Top / walking speed: 5 km/h
  
  
  

• Runtime per charge: ~2 hours
  
  
  

• Payload / strength: ~3 kg
  
  
  

• Hands / fingers: not available (narrative describes 12-DOF hands)
  
  
  

• Actuators: FSA 2.0 joints, up to ~380 N·m peak torque
  
  
  

• Safe with humans: not available
  
  
  

• Perception / OS / connectivity / LLM / IP: not specified
  
  
  

• Tooling / SDK: NVIDIA Isaac Lab, ROS, MuJoCo; VR teleop; lead-through; direct command
  
  
  

• Status: Prototype
  
  
  

• Verification: Verified (Humanoid Guide lists 5/10)
  
  
  

• Manufacturer: Fourier Intelligence
  
  
  

• Origin: China
  
  
  

• Website: https://www.fftai.com/products-gr2
  
  

• Markets / use cases: locomotion R&D, manipulation research, embodied-AI training, academic/industrial pilots
  
  
  

FAQ

1. Who is GR-2 for?
   It is aimed at labs, universities, and industrial R&D teams that need a human-scale testbed for loco-manipulation and embodied AI.
   
   
   

2. How is it better than GR-1?
   GR-2 adds more DOFs, stronger actuators, and a cleaner modular design, making it easier to maintain and more capable for complex skills.
   
   
   

3. Can I use ROS or Isaac with it?
   Yes. It supports ROS, MuJoCo, and NVIDIA Isaac Lab, plus VR teleop and lead-through modes.
   
   
   

4. How long does it run?
   About 2 hours per charge, which is typical for research sessions and demos.