NVIDIA Earth-2
Verified Oct 15, 2026 ActiveNVIDIA's open family of AI weather and climate models, covering forecasting, nowcasting, and high-resolution downscaling.
Listed on EffectHub ·
Pricing
- Pricing model
- 100% Free
The models and the Earth2Studio framework are published as open models and open software, free to self-host from GitHub and Hugging Face. Hosted use through NVIDIA NIM microservices follows NVIDIA's general enterprise software terms, and the product page states no price for that path.
Use tool ↗About NVIDIA Earth-2
Earth-2 is NVIDIA's family of open models, libraries and frameworks for weather and climate simulation rather than a single app. Its pieces are the Atlas architecture for medium-range forecasts of more than 70 variables up to 15 days out, a nowcasting model that generates satellite and radar imagery for zero-to-six-hour severe weather, a data-assimilation model that builds initial atmospheric state on GPUs in seconds, FourCastNet 3 for global forecasting, and CorrDiff, a generative downscaling model NVIDIA says runs 500 times faster than conventional methods. Earth2Studio on GitHub and the model collection on Hugging Face let teams fine-tune and run everything on their own hardware, and some models also ship as NIM microservices. It is aimed at meteorologists, climate researchers and companies building weather products, not at consumers.
- Platforms
- API
- API
- Yes
- Last verified
- Oct 15, 2026
Verified sources
- Described as open, production-ready AI models and tools for global forecasting: Atlas medium-range model for 70+ variables up to 15 days, nowcasting for 0-6 hour hazards, global data assimilation on GPUs, FourCastNet 3, CorrDiff downscaling claimed 500x faster; distributed via Earth2Studio on GitHub and a Hugging Face collection; no price stated (15 October 2026)nvidia.com ↗
- Earth2Studio, the open-source framework for building, fine-tuning and deploying Earth-2 models on your own infrastructuregithub.com ↗
Is NVIDIA Earth-2 legit?
We don’t label tools safe or unsafe. Below are the records we checked ourselves, each with the date we read it.
- Domain registered
- Apr 20, 1993 (33 years ago)Registry record (RDAP), registrar SafeNames Ltd., looked up Sep 6, 2026
- TLS certificate
- Issued by DigiCert Inc, valid until Dec 11, 2026Seen in our own connection to the site on Oct 15, 2026
- Last checked
- Oct 15, 2026The site answered
- Operated by
- NVIDIA Corporation, United StatesOur research, card verified Oct 15, 2026
Pros & cons of NVIDIA Earth-2
Earth-2 is a research-grade toolkit for weather and climate AI, not a polished self-serve product - its value is giving people building weather products access to models like CorrDiff and Atlas without training them from scratch. It requires real infrastructure and domain expertise to put to use; there is no simple sign-up-and-go path for a non-technical user.
Pros
- Covers the full pipeline from data assimilation through forecasting to high-resolution downscaling, not just one step
- CorrDiff claims large speedups over traditional downscaling approaches
- Open-source distribution via GitHub and Hugging Face, so it can be inspected and self-hosted
- Backed by NVIDIA's GPU computing expertise and ongoing development
Cons
- No pricing is disclosed for the hosted NIM microservice path
- Requires ML/infrastructure expertise to deploy - not usable by a non-technical audience
- As an open framework, it needs real GPU hardware to run at any useful scale
Best for
- Meteorology and climate research teams
- Companies building weather-dependent products who want a starting model instead of training one
- GPU-equipped teams wanting to self-host advanced downscaling or nowcasting models
Not for
- Anyone wanting a simple consumer weather-forecast app
- Teams without GPU infrastructure or ML engineering resources to run the models