WeatherNext 3, developed by Google DeepMind and Google Research, processes real-time satellite data to update global weather forecasts every hour. According to Google tests, the model provides up to a 50% increase in accuracy for precipitation forecasts one day or more ahead. WeatherNext 3 offers a resolution of 5 kilometers for surface variables such as temperature and humidity, 10 kilometers for some surface variables, and 25 kilometers for atmospheric variables such as wind speed. The previous WeatherNext 2 operated with a 25-kilometer grid and 6-hour intervals. The new system monitors sudden changes, precipitation and storms more closely, particularly in coastal and mountainous regions. The model also helps forecast renewable energy generation using wind speed, cloud cover and solar radiation data at 100 meters. The forecasts will be used in Google Search, Gemini, Google Maps, Google Maps Platform, Google Cloud and Earth Engine.
Why it matters
WeatherNext 3’s main distinction is not only that it updates weather forecasts more frequently, but also that it can track rapidly changing conditions in greater detail. This could provide earlier and more localized information for users making decisions based on the risk of precipitation or storms in coastal and mountainous regions. Adding data on wind, cloud cover and solar radiation to energy generation forecasts allows weather conditions to be used more directly in planning renewable energy resources. Extending the system to services ranging from Google Search to Cloud and Earth Engine means the forecasts will be visible not only in specialist tools but also on platforms used every day. The remaining question is how the improvement in accuracy reported in Google’s tests will translate across different regions and real-world conditions.
Background
Google is no stranger to the FikirPilot archive: over the past 90 days, we have published 15 news stories mentioning the name, the most recent dated September 4, 2026.