Google DeepMind open sources WeatherNext AI model for cyclone forecasting

Google DeepMind reports that its WeatherNext AI model has achieved state-of-the-art accuracy in predicting a cyclone’s track, intensity, and wind structure. According to the announcement, the model provides forecasters with an average of one extra day of predictive accuracy: three-day forecasts from WeatherNext are as accurate as two-day forecasts from prior models. The company characterizes this improvement as roughly equivalent to a decade of progress in meteorological modeling.

The research was conducted in collaboration with the National Hurricane Center (NHC), the Cooperative Institute for Research in the Atmosphere (CIRA), the UK Met Office, and weather agencies globally. During the 2025 hurricane season, WeatherNext contributed to a historic forecast for Hurricane Melissa by predicting its rapid intensification and landfall in Jamaica, enabling the NHC to issue an advance warning that gave local teams additional time to prepare.

WeatherNext uses a single AI model to bridge a longstanding forecasting gap: it jointly predicts global weather patterns and fine-scale cyclone dynamics up to 15 days in advance. The model generates localized probability maps of tropical storm to hurricane-force winds by running a 1,000-member ensemble. This approach avoids the traditional trade-off between coarse global models for track prediction and high-resolution local models for intensity, integrating both capabilities into one system.

To train WeatherNext, researchers co-trained on two data modalities: global weather dynamics and expert-curated historical cyclone observations. The team evaluated WeatherNext Cyclones on historical cyclones from 2023 to 2024, reporting an average lead-time advantage of more than 24 hours for predicting cyclone track, intensity, and wind structure compared to other top weather models.

Google DeepMind is now open sourcing the WeatherNext 2 and WeatherNext Cyclones models used during the 2025 hurricane season. The move aims to empower the research community and support local forecasters, renewable energy planning, and disaster preparedness with advanced AI tools.