Tech Desk: Google DeepMind has unveiled WeatherNext 3, an advanced AI weather model promising a revolution in forecasting accuracy. Moving beyond its predecessor, WeatherNext 3 leverages live satellite data, offering unprecedented hourly and high-resolution forecasts down to a 5-kilometer grid – a significant leap from the previous 25-kilometer standard. This immediate data integration eliminates the six-hour lag associated with traditional numerical weather prediction models, providing a continuously updated view of atmospheric conditions.

Crucially, WeatherNext 3’s enhanced precision extends to granular details like 100-meter wind speeds, vital for renewable energy companies optimizing wind farm output. It also offers superior cloud cover and sun radiation levels, benefiting solar energy production. By incorporating sparse weather station data, the model can predict highly localized variables such as humidity with remarkable accuracy, addressing variations that change rapidly over short distances.

Combining precipitation data from NASA and Google’s own satellite analysis leads to up to 50 percent more accurate rain and snow forecasts, particularly improving reliability in historically challenging regions. This means more dependable warnings and better planning for communities and industries.

WeatherNext 3 is set to power Google’s core weather experiences across Search, the Gemini app, and Google Maps, as well as the Google Maps Weather API and Google Earth Engine. For developers and enthusiasts, the original WeatherNext model has been open-source since August 2026, inviting further innovation. Experience the future of weather prediction today in the Google Weather Lab.