DeepMind’s WeatherNext 2: AI Model Forecasts Global Weather Two Weeks Ahead, 8X Faster

DeepMind’s WeatherNext 2 marks a significant upgrade in AI-powered weather forecasting, offering global forecasts up to two weeks ahead and covering temperature, atmospheric pressure, and wind patterns. According to DeepMind, the model performs computations eight times faster than its predecessor, enabling faster decision‑making for industries such as energy trading.

The defining enhancement is far more granular forecasting, including hourly predictions that can help organisations strengthen operational resilience against shifting conditions. WeatherNext 2 can now predict tropical storm trajectories, including hurricanes, with reliable accuracy up to three days in advance, extending the previous limit.

The improvement stems from a redesign of the model’s methodology, with WeatherNext 2 achieving results through a single processing step and reducing reliance on costly AI computing infrastructure. Ferran Alet notes that gaps in training data may still cause difficulties when predicting extreme rain or snow events, highlighting ongoing limitations despite the advances.