Unitree is teaming up with NVIDIA to make humanoid robot research a little less like assembling a science project from a thousand random parts.
The company announced the H2 Plus, a new humanoid robot reference design built on NVIDIA’s Isaac GR00T development platform. The goal is to give universities, labs, and robot developers a more complete starting point for humanoid research instead of forcing every team to stitch together hardware, simulation tools, training systems, sensors, and onboard compute from scratch.
The H2 Plus combines Unitree’s H2 humanoid chassis with Sharpa Wave tactile five-finger hands, NVIDIA Jetson Thor onboard computing, and Isaac GR00T software workflows. Unitree describes it as a unified system that gives researchers both the “body” and the “brain” needed to begin building robot skills and testing them in the real world.
That matters because humanoid robotics is still a messy field. A robot that can walk is one thing. A robot that can see, reason, use its hands, learn from demonstrations, train in simulation, and then perform tasks outside the lab is a much bigger problem. Unitree and NVIDIA are trying to package more of that pipeline into one reference platform.
The robot itself is human-scale. Unitree says the H2 Plus stands nearly 6 feet tall, weighs 150 pounds, and has 31 degrees of freedom across its body. Add the dual five-finger hands, and the system reaches 75 degrees of freedom across the body and hands. That gives researchers a machine built not just for walking demos, but for manipulation work where the robot has to reach, grip, lift, and interact with the world.
The H2 Plus also includes multiple cameras, an inertial measurement unit, microphones, speakers, Wi-Fi 6, Bluetooth 5.2, Ethernet, and USB connectivity. Its onboard NVIDIA Jetson AGX Thor T5000 includes a Blackwell GPU, 128GB of unified memory, and enough AI performance for real-time robot inference and control. The battery is rated for about three hours of operation.
NVIDIA CEO Jensen Huang framed humanoids as part of a much larger physical AI shift, saying humanoid robots could bring physical AI to major global industries and open a multitrillion-dollar economic opportunity. That is the big picture version. The practical version is that NVIDIA wants Isaac GR00T to become a standard toolkit for teaching robots how to do useful things.
The software side includes tools for teleoperation data capture, open GR00T foundation models, Isaac Sim and Isaac Lab for simulation and testing, Isaac ROS for deployment, and Jetson Thor for running models on the robot. Researchers can use the full stack or plug pieces into their own development workflow.
Unitree says the H2 Plus will be available in late 2026. The NVIDIA Isaac GR00T platform and reference workflow for Unitree’s smaller G1 humanoid are also expected to arrive on GitHub and Hugging Face.
For Clanks readers, the big takeaway is simple. Unitree and NVIDIA are not just showing off another humanoid robot. They are trying to make a standard research platform for the people who will teach these machines what to actually do.
Join the Discussion