NVIDIA's Isaac GR00T opens humanoid robotics research with 2,070 teraflops
Open-source humanoid robot design packs 2,070 FP4 teraflops for academic labs...
NVIDIA has announced the Isaac GR00T Reference Humanoid Robot, an open humanoid robot reference design built on the NVIDIA Jetson Thor platform and the Isaac GR00T open development environment. This initiative aims to democratize frontier humanoid robotics research by giving academic labs access to advanced hardware and a fully open software stack. The robot is powered by the Jetson AGX Thor T5000, delivering 2,070 FP4 teraflops of AI performance — enough to run real-time perception, planning, and control algorithms on board without relying on external compute.
By releasing the full design and software openly, NVIDIA enables researchers to focus on high-level algorithms for locomotion, dexterous manipulation, and autonomous navigation rather than building custom hardware from scratch. The Isaac GR00T ecosystem includes simulation tools, reinforcement learning frameworks, and pre-trained models for common tasks. This lowers the barrier to entry for labs that previously lacked the resources to build sophisticated humanoid robots, accelerating progress toward general-purpose humanoid robots for industrial and service applications.
- NVIDIA's Isaac GR00T is an open reference design built on Jetson Thor, targeting academic research
- Jetson AGX Thor T5000 delivers 2,070 FP4 teraflops for onboard AI processing
- Open software stack includes simulation tools, reinforcement learning frameworks, and pre-trained models
Why It Matters
Open-source humanoid hardware and software let more labs experiment with general-purpose robots, speeding up real-world deployment.