DeepMind’s GenCast beats ECMWF’s ensemble forecast on 97.2% of targets
See how this story connectsGoogle DeepMind published GenCast in Nature on 4 December 2024, a probabilistic machine-learning weather model that produces 15-day ensemble forecasts at 0.25° resolution. It was more accurate than ECMWF’s ENS ensemble on 97.2% of 1,320 tested targets, and on 99.8% at lead times beyond 36 hours, and one 15-day ensemble member takes about 8 minutes on a single TPU v5. DeepMind released the code and weights.
Why it matters
Weather services got an open ensemble model that outscored ECMWF’s operational ensemble on 97.2% of targets, less than three months before ECMWF put its own AI model into operation.
- When
- 4 Dec 2024
- ID
WTA-N-2024-0033- Where
- United Kingdom, International
- Who
- Google DeepMind, ECMWF
Summary by wheresthe.ai, from Google DeepMind and the reporting credited above. Checked against the source on 9 Oct 2026.
Source copy: download the page as saved on 18 Oct 2026 (PDF). Also held by the Internet Archive.
Cite this story
WTA-N-2024-0033: “DeepMind’s GenCast beats ECMWF’s ensemble forecast on 97.2% of targets.” wheresthe.ai, 4 Dec 2024. https://wheresthe.ai/d/deepmind-gencast-nature-2024-12/ (source: Google DeepMind).
Spotted an error? Email [email protected].