Google DeepMind and Google Research have developed WeatherNext, an AI model that significantly improves cyclone forecasting. Published in Nature, the model achieves state-of-the-art accuracy in predicting cyclone track, intensity, and wind structure, providing forecasters with an average of one extra day of predictive accuracy.

This advancement is equivalent to a decade of meteorological progress. The model was developed in collaboration with agencies like the National Hurricane Center (NHC) and the UK Met Office. During the 2025 hurricane season, WeatherNext aided the NHC in forecasting Hurricane Melissa, enabling crucial advance warnings. The models, WeatherNext 2 and WeatherNext Cyclones, are now being open-sourced to empower the research community and enhance disaster preparedness.

WeatherNext bridges the gap between global and local forecasting models by using a single AI system. It can predict a cyclone's path and strength up to 15 days in advance, generating localized probability maps. The model was trained on vast datasets of global atmospheric data and historical cyclone observations, enabling it to learn complex patterns and extreme weather behavior.