2026 · NVIDIA · Advanced
NVIDIA Isaac GR00T N1.7
GR00T N1.7 is an open vision-language-action model for generalized humanoid manipulation skills.
Direct answer
What does NVIDIA Isaac GR00T N1.7 contribute?
GR00T N1.7 is an open vision-language-action model for generalized humanoid manipulation skills.
Background
GR00T N1.7 takes multimodal inputs such as images and language and predicts actions for humanoid manipulation. NVIDIA presents it as a cross-embodiment model trained on bimanual, semi-humanoid, and humanoid data, with post-training for new robots and tasks.
Problem
The work addresses a central constraint in VLA: building systems that learn useful representations or actions while remaining general enough to transfer beyond a single demonstration or environment.
Core idea
GR00T N1.7 is an open vision-language-action model for generalized humanoid manipulation skills.
Architecture and method
GR00T N1.7 takes multimodal inputs such as images and language and predicts actions for humanoid manipulation. NVIDIA presents it as a cross-embodiment model trained on bimanual, semi-humanoid, and humanoid data, with post-training for new robots and tasks.
- Open humanoid VLA model
- Cross-embodiment robot data mixture
- Post-training path for specific robot platforms
Results and impact
It is one of the clearer bridges between research-grade VLA models and commercially usable humanoid robot policies.
Prerequisites
- VLA models
- Humanoid control
- Robot demonstrations
Recommended reading order
Read the explanation above, review the related topic pages, then use the primary-source links below to inspect the abstract, figures, experiments, and released implementation.
Primary sources
External links are provided after the context needed to evaluate the work.
Follow-up research
Related papers and concepts
RT-1: Robotics Transformer for Real-World Control at Scale
RT-1 trains one transformer policy on a large multi-task dataset of real robot demonstrations.
RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control
RT-2 co-trains vision-language models on web and robot data so semantic knowledge can influence actions.
Open X-Embodiment and RT-X
Open X-Embodiment combines robot datasets across institutions and trains policies that transfer across embodiments.
OpenVLA: An Open-Source Vision-Language-Action Model
OpenVLA is an open 7B-parameter VLA trained on the Open X-Embodiment dataset.
Common questions
Frequently asked questions
What is the main idea of NVIDIA Isaac GR00T N1.7?
GR00T N1.7 is an open vision-language-action model for generalized humanoid manipulation skills.
Why is NVIDIA Isaac GR00T N1.7 important?
It is one of the clearer bridges between research-grade VLA models and commercially usable humanoid robot policies.
What should I learn before reading NVIDIA Isaac GR00T N1.7?
Recommended prerequisites are VLA models, Humanoid control, Robot demonstrations.