Google DeepMind and Google Research have launched WeatherNext 2, a cutting-edge weather forecasting model that significantly improves prediction speed and resolution. This new model generates forecasts eight times faster than its predecessor and can provide hourly predictions. It uses advanced AI technology, allowing it to produce hundreds of potential weather scenarios from a single input. This is vital for planning, especially when considering worst-case scenarios that impact decisions in sectors like supply chains and aviation. WeatherNext 2 employs a Functional Generative Network (FGN) approach, which introduces variation while ensuring that forecasts remain physically realistic. It focuses on "marginals," which are standalone weather elements, and learns to forecast "joints," or complex interrelated systems. This method enables more nuanced predictions, such as regional heat waves or energy production at wind farms. The model outperforms the previous WeatherNext version on 99.9% of variables, including temperature and humidity, across lead times of 0-15 days. The forecast data from WeatherNext 2 is now accessible through Google’s Earth Engine and BigQuery, and an early access program is available on Google Cloud’s Vertex AI platform. This rollout aims to make advanced weather forecasting tools widely accessible, supporting researchers and businesses in tackling complex weather-related challenges. Google plans to continue improving the model by integrating new data sources and expanding its availability.