Google's GenCast AI Sets New Benchmark for Weather Forecasting

DeepMind's AI model achieves unprecedented accuracy in 15-day forecasts, outperforming the world's leading weather prediction system.

GenCast, developed by Google DeepMind, outperformed the European Centre for Medium-Range Weather Forecasts' (ECMWF) ENS system in 97.2% of tests comparing 2019 weather data.
Unlike traditional physics-based models, GenCast uses machine learning trained on four decades of historical data to generate probabilistic weather forecasts.
The AI model excels in predicting extreme weather events, such as tropical cyclone paths, and provides up to 12 additional hours of warning for severe storms.
GenCast operates with greater computational efficiency, producing 15-day forecasts in minutes compared to hours required by traditional supercomputer-based systems.
DeepMind has made GenCast's code open source and plans to release real-time and historical weather forecasts, enabling broader research and operational use.