Google released the WeatherNext 3 AI model for global weather forecasting
The new model provides hourly forecasts at five times the resolution of its predecessor by using live satellite data.

Google DeepMind and Google Research launched WeatherNext 3 to provide more precise weather predictions. Unlike traditional models that rely on physics simulations, this version analyzes real-time satellite imagery to deliver updates every hour. It offers a global view at a 5-kilometer resolution for temperature and moisture data.
The system is currently integrated into Google Search, Gemini, and Google Maps to assist with daily planning. It also provides specific metrics for renewable energy, such as wind speed at turbine height and solar radiation levels, to help energy companies manage power grids.
WeatherNext 3 brings us closer to a future where forecasts truly match what is happening on the ground.
While this model improves accuracy for rain and snow, the developers note that weather remains inherently unpredictable. Users should still consult local meteorological agencies for official warnings regarding extreme conditions or public safety threats.
Source: Google DeepMind Blog