Researchers from Google DeepMind and Google's in-house research team have developed an advanced AI weather forecasting model—WeatherNext 3. This model leverages a Functional Generative Network architecture, which enables it to perform computations around 8 times faster than its predecessor. Additionally, it excels at generating high-resolution weather forecasts at an hourly level. Over a 15-day forecast period, WeatherNext 3 surpasses its previous iteration in 99.9% of meteorological variables, including temperature, wind speed, and humidity. From a single initial state, it can infer hundreds of potential weather development scenarios, with each inference taking less than a minute on a single TPU. Presently, the model's forecast data is accessible on the Earth Engine and BigQuery platforms, and has been integrated into services such as Google Search, Gemini, Pixel Weather, and the Weather API of the Google Maps Platform.