25 Feb 2025

ECMWF puts AIFS machine-learning forecast model into operational use alongside physics model

Climate and environmentExperimental cross-disciplinary systems
AI illustration · Illustrative scene, not a photograph of the specific device or experiment.

What was demonstrated

On 25 February 2025 ECMWF began running its machine-learning model AIFS in operations, side by side with its physics-based Integrated Forecasting System (IFS). ECMWF reports that AIFS outperforms state-of-the-art physics-based models on many measures, including tropical cyclone tracks, with gains of up to 20%, and uses approximately 1,000 times less energy to make a forecast. The announcement does not name the baseline, the measure behind the 20% figure, or the verification period, and it gives no figure for speed.

Why it matters

ECMWF describes AIFS as the first fully operational open weather prediction model using machine learning, and the announcement presents it as part of the live forecasting service rather than only a research test. The performance figures are ECMWF's own, so the announcement shows operational use more clearly than it shows forecast skill.

Caveats

  • The performance and energy figures are ECMWF's own and are not independently verified in the announcement. No baseline, period or verification method is given for them.
  • The announcement does not say which measure the 20% gain applies to, so it cannot be read as a general improvement in forecast accuracy.
  • AIFS runs alongside the physics-based IFS rather than replacing it, so the operational result is a hybrid service and not a full substitution.

Technology profiles

Sources

  1. other
    ECMWF's AI forecasts become operational
    ECMWF, 25 Feb 2025.

Last reviewed 21 Oct 2026.