Google DeepMind, in collaboration with Google Research, has unveiled the WeatherNext 3 weather model, setting a new global benchmark for accuracy and real-time capabilities. By directly scrutinizing satellite observation data, this model delivers high-resolution weather forecasts that are refreshed on an hourly basis. It utilizes a dynamic data input system to swiftly capture and adapt to extreme weather fluctuations, boasting spatial resolutions of up to 5-kilometer precision for surface variables and 25-kilometer accuracy for atmospheric variables. Leveraging innovative training techniques, the model ensures that regions across Latin America, Africa, and the Asia-Pacific region enjoy weather forecasts that are just as precise as those in developed nations. In the realm of energy, its renewable energy prediction module furnishes vital data for wind farms and solar power plants, facilitating efficient grid management. Presently, the model is seamlessly integrated into the Google ecosystem, and its open weather API is poised to catalyze in-depth applications across a multitude of sectors.