NOAA AIGFS surface and pressure forecasts are now on GribStream

GribStream now supports NOAA/NCEP's operational Artificial Intelligence Global Forecast System deterministic forecast as two API datasets:

• AIGFS Surface, code aigfssfc
• AIGFS Pressure, code aigfspres

That means AIGFS can be queried through the same /timeseries, /runs, point, and grid workflows already used for GFS, GEFS, IFS, and GraphCast GFS.

What AIGFS is

NOAA made AIGFS v1.0 operational on December 17, 2025 at the 12Z cycle, alongside AIGEFS and HGEFS. The deterministic AIGFS model is based on Google DeepMind's GraphCast model and was developed by NCEP with NOAA research laboratories and EPIC.

The important operational detail is that this is not just a research chart product. It is a NOAA/NCEP operational global model feed published on NOMADS, with four cycles per day and forecast steps out to 384 hours.

What is available

The aigfssfc dataset is the compact near-surface feed:

• UGRD and VGRD at 10 m above ground
• TMP at 2 m above ground
• PRMSL at mean sea level
• APCP accumulated precipitation at the surface

The aigfspres dataset contains the upper-air pressure-level feed:

• HGT geopotential height
• TMP temperature
• SPFH specific humidity
• UGRD and VGRD wind components
• VVEL pressure vertical velocity

Pressure-level fields are available on 13 standard levels: 50, 100, 150, 200, 250, 300, 400, 500, 600, 700, 850, 925, and 1000 mb.

Both datasets are global, on a 0.25 degree latitude-longitude grid, with 00, 06, 12, and 18 UTC cycles and 6-hour forecast steps through hour 384.

Why this matters

AIGFS gives users another operational global deterministic forecast to compare against physics-based guidance. For practical workflows, it is useful when you want to ask questions like:

• Does the AI forecast agree with GFS on the synoptic pattern?
• Is AIGFS faster to update for a dashboard view that only needs a few surface fields?
• How different are AIGFS pressure-level winds or temperatures from GFS, IFS, or GraphCast GFS at the same valid time?
• Can an operational AI model provide a useful extra signal for medium-range decision support?

Because GribStream exposes AIGFS through the existing API shape, those comparisons do not require a separate NOMADS scraper in client code. You can request AIGFS, GFS, GEFS, IFS, and GraphCast GFS with the same JSON pattern and the same coordinate or grid workflows.

A practical comparison workflow

The useful way to evaluate AIGFS is side by side with the model families forecasters and forecast systems already depend on. For example, you can request TMP and 10 m wind from aigfssfc and compare the same coordinates and valid times against GFS. For upper-air pattern checks, compare HGT, TMP, UGRD, and VGRD from aigfspres against GFS, IFS, or GraphCast GFS.

That gives you a concrete AIGFS forecast API workflow: use the surface product for dashboard fields and customer-facing point forecasts, use the pressure product for synoptic context, and keep the same request structure when switching between AI and physics-based global models.

Start here

• AIGFS Surface model page
• AIGFS Pressure model page
• API docs

Sources

• NWS Service Change Notice 25-89, "Implementation of the Artificial Intelligence Global Forecast System (AIGFS), Artificial Intelligence Global Ensemble Forecast System (AIGEFS), and the Hybrid Global Ensemble Forecast System (HGEFS)": https://www.weather.gov/media/notification/pdf_2025/scn25-89_AIGFS_AIGEFS_and_HGEFS.pdf
• NCEP AIGFS product inventory: https://www.nco.ncep.noaa.gov/pmb/products/aigfs/
• NOAA NOMADS AIGFS operational data path: https://nomads.ncep.noaa.gov/pub/data/nccf/com/aigfs/prod/
• NOAA/NCEP Office Note 521: https://doi.org/10.25923/xd3y-wy31