NVIDIA launches Physical AI Open Datasets: 7M trajectories, 1K assets
Earlier this year NVIDIA dropped a new open-source AI bundle on Hugging Face - the NVIDIA Physical AI Open Datasets. Inside you’ll find more than 7 million robotics trajectories plus about 1,000 OpenUSD SimReady assets. It’s a mix: some data comes from real-world captures, the rest is synthetic, pulled from NVIDIA’s Cosmos, Isaac and DRIV platforms.
Since the launch the package has been downloaded over 6 million times, so there’s clearly a lot of curiosity from researchers. The release was timed with Open Source AI Week, an event that tries to showcase how developers are pushing AI forward. The sheer size of the trajectory set looks promising for training and testing physical-AI models, yet I’m not sure how smoothly the synthetic pieces will mesh with the real ones in practice.
The datasets are free to access, but the announcement didn’t spell out licensing details, quality scores or benchmark results. In short, the potential is there, but we’ll have to wait and see how useful it really becomes.
NVIDIA earlier this year released the NVIDIA Physical AI Open Datasets on Hugging Face, featuring more than 7 million robotics trajectories and 1,000 OpenUSD SimReady assets. Downloaded more than 6 million times, the datasets combines real‑world and synthetic data from the NVIDIA Cosmos, Isaac, DRIVE and Metropolis platforms to kickstart physical AI development. NVIDIA Inception Startups Highlight AI Innovation 🔗 At the PyTorch Conference’s Startup Showcase, 11 startups — including members from the NVIDIA Inception program — are sharing their work developing practical AI applications and connecting with investors, potential customers and peers.
The week wrapped up on a high note, but the real test for NVIDIA’s Physical AI Open Datasets is probably more than just the download count. Six million-plus pulls hint at curiosity, yet it’s still unclear whether developers will turn those 7 million trajectories and 1,000 OpenUSD SimReady assets into working robotics pipelines. The collection lives on Hugging Face and blends real-world and synthetic data from NVIDIA’s Cosmos, Isaac and DRIV platforms - a mix that could smooth simulation-to-real workflows, if the integration hurdles aren’t too steep.
Open-source fans cheered the effort as a concrete outcome of the workshops and talks that dotted Open Source AI Week, pointing out the community’s willingness to tinker with fresh data sources. But no solid examples of downstream use showed up, leaving the practical impact a bit vague. We’ll likely hear more at NVIDIA GTC in Washington, D.C., where future sessions might reveal adoption trends.
For now, the datasets sit as a promising, yet still unproven, piece of the larger push toward accessible physical AI.
More in Open Source
Anthropic CEO Dario Amodei calls AGI a marketing term, echoing Altman's view
The AI community has grown weary of a word that shows up on every press release and investor deck.
Xiaomi launches MiMo‑V2‑Flash AI model: 150 t/s, USD 0.1‑USD 0.3 per million tokens
Xiaomi’s latest foray into generative AI arrives as an open‑source model named MiMo‑V2‑Flash, a...
OpenAI signs USD 10 bn deal to use Amazon Trainium3 chips, 4.4× faster compute
OpenAI has just inked a ten‑billion‑dollar agreement to run its models on Amazon’s newest Trainium3...
Zepto Cafe uses MCP to parse text orders, Playwright runs the clicks
Zepto Cafe has taken a modest step toward turning plain‑text requests into a full‑fledged ordering...
Larian CEO denies AI cuts, says generative AI hasn't boosted efficiency
Why does the Larian story matter now? The studio’s founder‑CEO, Swen Vincke, has been under fire...
Common Questions Answered
What platforms contributed data to the NVIDIA Physical AI Open Datasets?
The datasets combine real-world and synthetic data sourced from several of NVIDIA's key platforms, including Cosmos, Isaac, and DRIVE. This diverse data foundation is intended to provide a robust starting point for developing physical AI applications.
How many times have the NVIDIA Physical AI Open Datasets been downloaded since their release?
According to the article, the data package has been downloaded over 6 million times since its launch. This high number indicates significant interest and curiosity from the research and developer community.
What is the composition of the data in the NVIDIA Physical AI Open Datasets collection?
The collection bundles more than 7 million robotics trajectories along with 1,000 OpenUSD SimReady assets. This blend includes both real-world captures and synthetic scenes, offering a comprehensive resource for simulation and development.
Where are the NVIDIA Physical AI Open Datasets hosted for public access?
NVIDIA has made these datasets available on the Hugging Face platform. Hosting them on this popular site facilitates easy access for developers and researchers worldwide who are interested in physical AI.