AI GFS - AIGFS, Artificial Intelligence Global Forecast System
| Provider: | National Centers for Environmental Prediction, NOAA (USA) |
| Model scope: | Global |
| Update frequency: | every 6 hours |
| Resolution: | 0.25°, 15.0nm, 27.8km |
| Model duration: | 49 forecasts starting at 0 hr, ending at 16 days |
| Parameters: | pressure, wind, rain, temperature, vertical velocity, 250 mb, 500 mb, 850 mb |
| GRIB model date: | Tue Feb 10 18:00:00 2026 UTC |
| Download date: | Tue Feb 10 22:32:40 2026 UTC |
| Download delay: | 4hr 32min |
Note: the Download delay is the amount of time required for the GRIB model to compute its forecast and then for the LuckGrib cluster to download the data and make it available. The LuckGrib delay is generally less than 10 minutes, the remainder of the delay is the model compute time.
Introduction
The U.S. National Weather Service (NWS) and National Centers for Environmental Prediction (NCEP) are now running an operational global forecast model based on machine learning (ML) and artificial intelligence (AI) techniques: AIGFS.
For background on how ML is being applied to numerical weather prediction, see the short discussion here.
AIGFS produces forecasts four times daily (00Z, 06Z, 12Z, 18Z), with output every 6 hours out to 16 days.
AIGFS details
AIGFS was officially announced in an NWS service change notice (SCN 25-89) with operational rollout in mid December 2025. For deeper technical background, see the NOAA repository document here and the underlying research paper here.
A key finding from the paper:
… Evaluation of the GCGFS forecast skills showed that both the trained and fine-tuned GCGFS models outperform the operational GFSv16 forecasts, especially at longer lead times.
We are still very early in the evolution of ML-based weather models, yet they are already delivering results comparable to, or in some cases better than, traditional physics-based systems like GFS.
Current limitations
AIGFS currently offers a smaller subset of parameters than the deterministic GFS. Notable missing fields include:
- Wind gusts
- Simulated radar
- Ocean waves and swell
- Visibility
- Cloud cover fractions
- Vorticity and divergence diagnostics
- Many others
These gaps are expected to narrow as the models mature, research is active in this area.