Skip to main content
UFFDA
WeatherParked

AIFS / GraphCast / GenCast (open AI weather ag-feed)

Foundation-model weather forecasts — AIFS (ECMWF's AI weather model), GraphCast, GenCast. Not run today; GPU inference is the cost.

What this is

AI-foundation-model weather feeds — DeepMind's GraphCast, ECMWF's AIFS, and GenCast — repackaged for agricultural use. All three are open-weight models; the barrier isn't licensing, it's the GPU compute needed to run inference at scale.

Why we surface it

These models are trained end-to-end rather than built from physics equations, and they're matching or beating traditional numerical models on several benchmarks. If UFFDA ever needs sharper short-range forecasts than NOAA's public models provide, this is where that would come from.

What you can do with it

Shown for commercial use. Personal, research, and nonprofit use is often more permissive — switch the verdict lens on the catalog.

Internal / individual use
Custom or restrictive terms — we can't auto-decode this one. Read the source agreement before using, especially for commercial use, AI training, and redistribution. When in doubt, ask the provider.
Redistribute the data
Custom or restrictive terms — we can't auto-decode this one. Read the source agreement before using, especially for commercial use, AI training, and redistribution. When in doubt, ask the provider.
Republish a derivative
Custom or restrictive terms — we can't auto-decode this one. Read the source agreement before using, especially for commercial use, AI training, and redistribution. When in doubt, ask the provider.
Build into a private product
Custom or restrictive terms — we can't auto-decode this one. Read the source agreement before using, especially for commercial use, AI training, and redistribution. When in doubt, ask the provider.

This decoder is a starting point, not legal advice. For commercial deployment, run the actual license terms past someone who reads them for a living.

Parked

Parked for now — running these models costs real money, roughly $50K a year in GPU time, and that's not justified without a clear integration need.

Related — same domain