Google DeepMind Launches WeatherNext 3: AI Weather Model Delivers Hourly Forecasts at 5-Kilometer Resolution Across Google Platforms
Summary
Google DeepMind launches WeatherNext 3, a groundbreaking AI weather model delivering hourly forecasts at 5-kilometer resolution — five times sharper than its predecessor — now powering Google Search, Maps, and Gemini while offering up to 60% improved precipitation accuracy and specialized renewable energy forecasts for clean energy grid planning.
Key Points
- Google DeepMind and Google Research launch WeatherNext 3, the most advanced global AI weather model to date, delivering hourly forecasts at up to 5-kilometer resolution — five times sharper than its predecessor — by ingesting real-time geostationary satellite data instead of relying on traditional physics-based simulations.
- WeatherNext 3 achieves breakthrough precipitation forecasting accuracy, improving Continuous Ranked Probability Scores by up to 60% against NASA's IMERG baseline, while also introducing specialized forecasts for renewable energy including 100-meter wind speeds and high-resolution solar radiation data to support clean energy grid planning.
- Starting today, WeatherNext 3 powers weather experiences across Google Search, Gemini, Google Maps, Google Maps Platform Weather API, and Google Earth Engine, with forecast data also available for developers and researchers via BigQuery, Earth Engine, and Google Cloud Storage.