Google WeatherNext 3 Moves AI Forecasts Into Products
The new global model produces hourly forecasts from satellite and station data and is rolling out across Search, Maps, Gemini and Google Cloud services.
Hourly global forecasting
Google DeepMind and Google Research have introduced WeatherNext 3, a global AI weather model designed to produce a new forecast every hour rather than waiting for the six-hour update cycle common to traditional global systems. Its forecasts began rolling out globally across Search, the Gemini app, Google Maps, Google Maps Platform and Earth Engine on September 3. Operational data is also accessible through BigQuery and Earth Engine or as bulk downloads from Cloud Storage.
The model ingests live geostationary satellite mosaics alongside traditional historical analysis and is also trained directly on sparse weather-station observations. Google says this lets it respond more quickly to changing rain and snow patterns while representing local conditions that can be poorly covered by conventional forecasting systems. WeatherNext 3 produces temperature and moisture fields at five-kilometer resolution, other surface variables at ten kilometers and atmospheric variables such as wind speed at 25 kilometers. It can also predict station-level locations directly.
Its design targets commercial forecasting as well. Outputs include turbine-height wind speeds, cloud cover and solar radiation relevant to renewable-energy operators, while hourly forecast initialization could help logistics, agriculture and utilities react to rapidly developing conditions. Google says the model improves precipitation detail and boundary definition, two areas where probabilistic AI forecasts have often appeared blurred.
Deployment at consumer scale
The most important change is operational rather than purely experimental. Google has begun placing the model behind products used by billions of people, creating a direct route from AI meteorology research to everyday decisions. Developers and organizations can access the same forecast family without installing or maintaining the model themselves.
Google cautions that WeatherNext remains an input to decision-making, not a replacement for official warnings from national and local meteorological agencies. Although Google cites independent live evaluations by Brightband, much of the detailed public evidence still comes from Google, so performance across rare events, poorly observed regions and different forecast horizons will need continued outside examination.
Why it matters
Weather forecasting is a consequential test of AI outside conversational products: predictions can be scored against physical reality and carry direct economic and safety effects. Hourly global updates combined with mass-market distribution could make WeatherNext 3 one of the most widely used scientific AI systems, provided its local accuracy holds up in live operations.