Google DeepMind launches WeatherNext Cyclones for unified track and intensity forecasting
Google DeepMind has released WeatherNext Cyclones, an AI system that jointly forecasts tropical cyclone tracks and intensity rather than treating them as separate problems. Reporting tied to a Nature publication indicates the model extends useful lead time by roughly one day compared with leading operational baselines, a gain likened to a decade of conventional forecasting progress. The Decoder notes that open-source code and weights accompany the announcement, following live trials with the U.S. National Hurricane Center on Google Weather Lab since June 2025. Coverage also highlights that WeatherNext can operate on lower-resolution inputs while still predicting path and strength, lowering deployment cost. Operators should treat the public release as an integration opportunity, but independent verification of real-season performance beyond cited trials remains limited.
Google DeepMind launches WeatherNext Cyclones for unified track and intensity forecasting
Google DeepMind is introducing WeatherNext Cyclones (WN-C), an AI system that forecasts tropical cyclone tracks and intensity together and looks about one day further ahead than leading operational models. The Decoder reports the Nature paper release ships open-source code and weights on GitHub after live runs with the NHC on Google Weather Lab since June 2025.
Key takeaway
Open-source cyclone track-and-intensity forecasting with about one extra day of lead time is now available for builders to integrate directly.
What happened
Google DeepMind introduced WeatherNext Cyclones (WN-C), an AI system that forecasts tropical cyclone tracks and intensity together. The Decoder reports the Nature paper release includes open-source code and weights on GitHub.
The Decoder says WN-C appears to extend forecasts about one day further ahead than leading operational models, after live runs with the National Hurricane Center on Google Weather Lab since June 2025. Wired reporting cited by Techmeme adds that the model can predict storm track and intensity from lower-resolution weather data.
Evidence
WeatherNext Cyclones jointly forecasts tropical cyclone tracks and intensity.
The Decoder · attributed
Google DeepMind is introducing WeatherNext Cyclones (WN-C), an AI system that forecasts tropical cyclone tracks and intensity together
The model extends forecasts about one day further ahead than leading operational models.
The Decoder · attributed
looks about one day further ahead than leading operational models
Open-source code and weights ship with the Nature paper release.
The Decoder · attributed
The Decoder reports the Nature paper release ships open-source code and weights on GitHub
Live runs with the NHC on Google Weather Lab began in June 2025.
The Decoder · attributed
after live runs with the NHC on Google Weather Lab since June 2025
WeatherNext can predict storm track and intensity using lower-resolution weather data.
Techmeme · attributed
Google DeepMind says its WeatherNext model can accurately predict a storm's track and intensity using lower-resolution weather data
DeepMind open-sourced WeatherNext 2 and WeatherNext Cyclones models.
Hacker News AI · attributed
DeepMind has open-sourced its WeatherNext 2 and WeatherNext Cyclones models, which provide accurate global weather forecasts and an extra day of warning for cyclones
Why it matters
Joint track-and-intensity cyclone forecasts with open weights lower the barrier for disaster-response, insurance, and logistics products that need earlier, unified storm intelligence.
Limits and uncertainties
Packet cites NHC live runs since June 2025 but does not provide independent benchmark results beyond comparisons to leading operational models.
Practical implications
Builders can pull open-source WeatherNext Cyclones code and weights from GitHub instead of licensing proprietary operational forecast feeds.
What to watch
Whether operational agencies beyond the cited NHC trials adopt WN-C outputs during the current hurricane season.
