Introducing WeatherNext 3, our most advanced and accurate global weather AI model

Weather affects billions of daily decisions, from carrying an umbrella to managing agriculture, supply chains, clean energy, and national economies. While AI has already improved weather forecasting, previous models lacked the spatial resolution and real-time data ingestion needed to predict highly local, rapidly changing weather such as short-lived storms or heatwaves.

WeatherNext 3 addresses this by learning directly from real-time observations. The end-to-end system ingests live 1-hour geostationary satellite mosaics alongside traditional historical analysis into a single, flexible Functional Generative Network (FGN) mesh transformer. It generates hourly forecasts at multiple spatial resolutions, producing dense gridded fields, discrete cyclone tracks, and station-level sparse coordinates natively. Temperature and moisture are resolved at 5 km, other surface variables at 10 km, and atmospheric variables like wind speed at 25 km.

According to independent live evaluations by Brightband, WeatherNext 3 is the most advanced and accurate global weather model to date, with a global picture about five times sharper than WeatherNext 2's 25 km, 6-hour grid. The model also adds precise precipitation forecasting and clean energy variables, and can update forecasts every hour using raw satellite data.

WeatherNext 3 is now integrated across Search, Gemini, Maps, Google Maps Platform, and Cloud, making its high-resolution forecasts accessible worldwide in Google products. This should improve localized forecast quality and support decisions in agriculture, logistics, and clean energy operations.