Google DeepMind Unveils WeatherNext Cyclones With Added Forecast Lead Time
The model's three-day forecast accuracy matches what older models delivered at two days, a gain the source equates to a decade of meteorological progress, and the open-sourcing of both models hands that capability to the research community.
Reporting from 1 source: GIGAZINE.
Google DeepMind announced WeatherNext Cyclones, an AI weather model developed with the US National Hurricane Center, CIRA, the UK Met Office, and other agencies. It achieves state-of-the-art accuracy in tracking tropical cyclones, offering forecast information at three days out that matches the accuracy of two-day forecasts from previous models. The models WeatherNext 2 and WeatherNext Cyclones are being open-sourced.
The new model builds on WeatherNext 2, released in November 2025, which enabled hourly high-resolution forecasts. WeatherNext Cyclones focuses specifically on the track, intensity, and wind structure of tropical cyclones, the systems known as typhoons, hurricanes, or cyclones depending on the ocean basin.
During the 2025 hurricane season, the system predicted the strengthening of Hurricane Melissa and its landfall in Jamaica, giving the National Hurricane Center time to issue early warnings. The collaboration continues into 2026, running 1,000 scenarios per storm to support forecaster decisions.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.