Home / Technology / AI Weather Model Predicts Future With Unprecedented Detail
AI Weather Model Predicts Future With Unprecedented Detail
4 Sep
Summary
- New AI model offers hourly forecasts in 5km grids, up from 25km.
- Incorporates live satellite data, overcoming numerical model lag.
- AI improves precipitation forecasts by up to 50% globally.
Google DeepMind has unveiled WeatherNext 3, an advanced AI weather model delivering unprecedented forecast detail. This updated system offers hourly predictions across a refined 5-kilometer grid, a substantial improvement over the previous 25-kilometer resolution.
The model's enhanced accuracy stems from its training on live satellite data. This approach bypasses the typical six-hour lag found in numerical weather prediction models, providing a continuous, real-time view of atmospheric conditions.
WeatherNext 3 is particularly valuable for the renewable energy sector. It can forecast wind speeds at turbine height, predict solar radiation, and estimate cloud cover with high precision, aiding solar and wind farms.
Furthermore, the model significantly improves precipitation forecasts, achieving up to a 50% increase in accuracy. This advancement is most pronounced in regions where weather predictions have historically faced challenges.
Users can experience WeatherNext 3 through Google Search, the Gemini app, and Google Maps. The model also powers the Google Maps Weather API and Google Earth Engine, with the original WeatherNext model available as open-source since August 2026.
