Google has launched WeatherNext 2, its most advanced AI weather forecasting model, delivering faster and more detailed global weather predictions than previous systems. The new model generates hundreds of possible weather scenarios in under a minute using just one TPU, while offering resolution down to the hour. This represents a significant leap in weather prediction capabilities.

WeatherNext 2 surpasses Google's previous WeatherNext model on 99.9% of variables and lead times. The breakthrough comes from a new AI approach called Functional Generative Network, which injects noise directly into the model architecture to maintain physical realism. This allows the system to predict both individual weather elements and complex interconnected systems.

The technology is now accessible through Google's Earth Engine and BigQuery, with an early access program on Vertex AI. WeatherNext 2 already powers upgraded forecasts in Search, Gemini, Pixel Weather, and Maps Platform. This AI-driven approach transforms weather forecasting from single deterministic predictions to comprehensive scenario analysis.