WeatherNext: AI model achieves breakthrough in forecasting cyclones
deepmind ·
Google DeepMind and international meteorological partners have introduced WeatherNext, an artificial intelligence system that markedly enhances tropical cyclone prediction capabilities. By accurately forecasting storm tracks, intensity, and wind structures, the model provides roughly one additional day of reliable lead time compared to existing methods. This performance jump equates to approximately a decade of traditional meteorological progress. During the 2025 hurricane season, the technology notably aided the National Hurricane Center in correctly predicting Hurricane Melissa’s rapid intensification and Jamaican landfall. In a move to accelerate global research and operational forecasting, the developers have fully open-sourced the WeatherNext Cyclones and WeatherNext 2 models alongside their underlying code. Built on twenty terabytes of atmospheric data and historical storm databases, the system employs functional generative networks to rapidly produce thousands of probabilistic outcome scenarios. Remarkably, the AI maintains high precision even when processing spatially coarse input data that falls far below standard industry resolutions. These open-access tools are now available through Weather Lab, offering scientists and emergency planners advanced resources to improve disaster readiness and community resilience.
deepmind