The AIFS (Artificial Intelligence Forecasting System) is ECMWF's machine-learning forecasting system. Its real-time output is published as open data, free of charge, under the Creative Commons CC-BY-4.0 licence and the ECMWF Terms of Use, with appropriate attribution. Data are released as soon as they are produced and kept for up to four days on the ECMWF open-data platform (longer on some mirrors).

The two AIFS models

ModelRequest keywordWhat it isData from
AIFS Singlemodel=aifs-singleDeterministic machine-learning forecastFebruary 2025
AIFS ENSmodel=aifs-ens (aifs-ens-crps in MARS)Ensemble machine-learning forecastJuly 2025

What's available

Both models are on a 0.25° global grid in GRIB edition 2, with 6-hourly steps out to 15 days (360h), on the same dissemination schedule as the IFS open data.

ModelStreamTypeWhat it isFormat
aifs-singleoperfcDeterministic forecastGRIB2
aifs-ensenfopf (use cf for the control member; ep for probabilities)Ensemble forecastGRIB2

Parameters are listed in the ECMWF shopping-cart tool under Group D: AIFS; see the AIFS machine-learning dataset pages for the full product list.

AIFS products use a different file-naming convention from the IFS. Build AIFS requests with the ecmwf-opendata package (below), or follow the AIFS naming reference in the open-data documentation, rather than copying the IFS pattern.

How to access it

AIFS open data is available from the same locations as the IFS open data (see Getting the data). With the Python package, set the model explicitly.

AIFS Single (deterministic):

from ecmwf.opendata import Client

client = Client(source="ecmwf", model="aifs-single")
client.retrieve(
    stream="oper", type="fc", param="2t",
    step=24, target="aifs-2t-step24.grib2",
)

AIFS ENS (ensemble — one perturbed member of mean sea-level pressure):

from ecmwf.opendata import Client

client = Client(source="ecmwf", model="aifs-ens")
client.retrieve(
    time=0, stream="enfo", type="pf", number=1,
    param="msl", target="aifs-ens-msl.grib2",
)

Other access routes

  • Dissemination (Gold and Silver service packs): add AIFS to a data flow in the Product Requirements Editor with class=ai and model=aifs-single or model=aifs-ens.
  • Archive / MARS: historical AIFS data via the model keyword (aifs-single; aifs-ens-crps for the ensemble). AIFS Single from February 2025, AIFS ENS from July 2025.
  • Charts: search for "aifs" in OpenCharts and ecCharts.
  • No labels