Unitree G1

Go Beyond the Limits

The Unitree G1 is the world's most popular and most published-on commercial humanoid robot — a compact, research-grade platform that democratized bipedal humanoid robotics. Trusted by institutions including Caltech, Cornell, MIT, and Stanford, the G1 delivers extraordinary capability in a lightweight, portable form factor, with configurations scaling from out-of-the-box locomotion to full dexterous manipulation research.

Ideal for: Academic and university research (locomotion, manipulation, whole-body control), reinforcement learning development, STEM education programs, robotics competitions, medical task prototyping, and commercial AI demos.

Technical Specifications

Physical

• Height (standing): 132 cm
• Weight (with battery): ~35 kg
• Materials: High-strength aluminum alloy + industrial-grade crossed roller bearings

Degrees of Freedom

• G1 Standard: 23 DOF (6 per leg, 5 per arm, 1 waist)
• G1 EDU: 23–43 DOF (expandable with additional waist axes, wrist axes, and dexterous hands)
• Optional Dex3-1 three-finger dexterous hand: 7 active DOF per hand (thumb 3 + index 2 + middle 2), with optional tactile sensor arrays

Performance

• Walking speed: up to 2 m/s
• Maximum knee joint torque: 90 N·m (standard) / 120 N·m (EDU)
• Battery: 9000 mAh quick-release smart pack, ~2 hours runtime, 30-second hot-swap
• World records: 1.4 m long jump, side flip, kip-up from ground

Computing & Intelligence

• G1 Standard: 8-core high-performance CPU with intelligent OTA upgrades
• G1 EDU: Dual-computer architecture — a dedicated real-time motion control computer handles balance and actuation at high frequency, while an open secondary development computer (NVIDIA Jetson Orin NX, 100 TOPS) is fully accessible via SSH for custom software, SDK development, and AI workloads
• UnifoLM-VLA-0 (open-source vision-language-action model) deployable on G1 EDU

Perception

• 3D LiDAR: LIVOX MID-360
• Depth cameras: Intel RealSense D435i
• IMU
• Joint encoder feedback
• Built-in stereo speakers and microphone array

Connectivity & Software

• ROS / ROS2 compatible (EDU)
• Full SDK access: C++, Python, ROS2 (EDU)
• Reinforcement learning support
• Remote control interface
• Motion scripting and trajectory planning
• Simulation environment compatibility (NVIDIA Isaac Gym)
• Wi-Fi 6 and Bluetooth 5.2