**[Google Weather AI Accelerates Again: 5-Kilometer Forecasts, Updated Hourly]** Beating AI News: Google DeepMind and Google Research have released WeatherNext 3, which directly integrates global real-time satellite observations to regenerate global weather forecasts every hour. The previous generation, WeatherNext 2, primarily used a 25-kilometer grid and updated every six hours. The new model achieves surface variables like temperature and humidity at a resolution of approximately 5 kilometers. Precipitation, including rain and snow, is a key area of improvement this time. WeatherNext 3 incorporates NASA satellite precipitation data and radar data for training. Google claims that in scenarios forecasting one day or further ahead, users can see precipitation forecast accuracy improve by up to 50%. Previous AI weather models often predicted localized heavy rainfall too vaguely, and this update focuses on enhancing the details of rapidly changing weather conditions. WeatherNext 3 has already been integrated into Google Search, Gemini, Maps, Google Maps Platform Weather API, and Earth Engine. Developers can also access forecast data via BigQuery, Earth Engine, and Google Cloud Storage. Traditional physical weather models will not disappear for now. WeatherNext 3's training still utilizes data from these models, and meteorological agencies will continue to combine multiple forecasts when issuing disaster warnings. Google also reminds users that severe weather and public safety alerts should still rely on local meteorological departments. [Original Link]