Google launches WeatherNext 2 for faster forecasts

Forecast data from WeatherNext 2 is now accessible via Earth Engine, BigQuery, Vertex AI, and integrated into Search, Pixel Weather, Gemini, and Google Maps.

Google’s WeatherNext 2 uses AI to generate hundreds of weather scenarios quickly, offering hourly-resolution forecasts and helping agencies plan for extreme events and everyday decisions.

WeatherNext 2, Google’s latest AI forecasting model, offers significantly faster and more precise weather predictions. Developed by DeepMind and Google Research, the model produces forecasts eight times faster with hourly resolution, aiding decisions from supply chains to daily commutes.

The model generates hundreds of weather scenarios from a single starting point, enabling agencies and businesses to plan for all potential outcomes, including extreme events.

Its predictions outperform the previous WeatherNext model on 99.9% of variables, providing more accurate forecasts for temperature, wind, humidity, and other factors.

A Functional Generative Network (FGN) powers WeatherNext 2, allowing it to predict both individual weather elements and complex interconnected systems. The system enables applications such as forecasting regional heatwaves or wind farm output, while keeping predictions physically realistic.

Forecast data is available through Google Earth Engine, BigQuery, and an early access programme on Vertex AI, while WeatherNext 2 now powers Search, Gemini, Pixel Weather, and Google Maps’ Weather API.

Google plans to expand access further, supporting researchers, developers, and businesses to make informed decisions and accelerate scientific discovery.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!