Tier 2 Restricted open-weight
What the publisher claims: open foundation model, open code
The one-liner: Open robot foundation model whose first weight releases are non-commercial.
The six questions
Code and checkpoints are downloadable from GitHub and the Hub, ungated.
The N1 and N1.5 weights carry a non-commercial licence; only the newer N1.7 weights moved to commercial-use terms.
Inference and fine-tuning run on NVIDIA GPUs, Jetson Thor or DGX Spark, with TensorRT export supported.
Technical report, architecture and evaluation protocol are published, but the pretraining data mixture is only partly released as a simulation dataset.
Fine-tuning and post-training are reproducible from the released code; pretraining from scratch is not.
NVIDIA's Isaac-GR00T repo takes issues; roadmap and releases are wholly NVIDIA's.
License
Code Apache 2.0; N1 and N1.5 weights under the NVIDIA one-way non-commercial licence per their model cards; newer N1.7 weights under the NVIDIA Open Model License — license source ↗
- N1 and N1.5 weights are non-commercial
- NVIDIA Open Model License requires an attribution notice and Cosmos branding for Cosmos-derived work
What is actually released
- reference code
- N1, N1.5 and N1.7 checkpoints
- PhysicalAI-Robotics-GR00T-X-Embodiment-Sim dataset
- fine-tuning and evaluation scripts
Why it matters
Robotics repeats the text-LLM pattern: open code and checkpoints, with permission lagging capability.
Sources
- https://huggingface.co/nvidia/GR00T-N1-2B
- https://huggingface.co/nvidia/GR00T-N1.5-3B
- https://raw.githubusercontent.com/NVIDIA/Isaac-GR00T/main/README.md
- https://raw.githubusercontent.com/NVIDIA/Isaac-GR00T/n1d5/README.md
- https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/
Derived from the openness dataset behind But Is It Open? — snapshot 2026-11-07 ↗. Scores are qualitative readings of the cited evidence, not verdicts; tier measures openness, not intelligence.