Google DeepMind Unveils WeatherNext 3 AI Weather Model Using Live Satellite Data

Google DeepMind and Google Research have unveiled WeatherNext 3, a new artificial intelligence weather model that uses live satellite data to deliver more frequent and higher-resolution forecasts. The system, announced on September 3, 2026, represents a significant shift from traditional methods that rely on data from physics-based simulations, which can be hours out of date. WeatherNext 3 ingests real-time imagery from geostationary satellites, allowing it to produce hourly updates instead of the standard six-hour intervals. This fresh data stream enables the model to resolve weather conditions at a 5-kilometer scale for key variables like temperature and moisture, a fivefold improvement over its predecessor, WeatherNext 2, which operated on a 25-kilometer grid. The model's architecture is designed to maintain physical consistency across global and local scales, helping it handle rapid changes in conditions like rain or wind. Independent testing on the Operational WeatherBench platform has validated the model's performance, ranking it as the most accurate global weather model currently available under real-time conditions. A particular strength is precipitation forecasting.