Google Unveils WeatherNext 3: Real-Time AI Model for Accurate Atmospheric Forecasts

Google has unveiled WeatherNext 3, its latest AI model for delivering fast and accurate atmospheric forecasts. This next-generation system is designed to provide real-time data on weather conditions across various Google platforms, including Search, Maps, Gemini, and Earth Engine.

The new model uses raw satellite feeds paired with historical data patterns to predict weather conditions down to a 5-kilometer resolution. Unlike its predecessor, WeatherNext 2, which updated every six hours, WeatherNext 3 processes fresh satellite observations every hour.

This improvement allows for more precise forecasts of temperature, wind speed, humidity, and moisture levels. The model also features 2.4 times more parameters than WeatherNext 2, enabling it to visualize complex patterns such as cyclone paths with greater accuracy.

Google reports that WeatherNext 3 improves precipitation accuracy by up to 50% for 24-hour lookaheads and boosts overall rain accuracy by 60% compared to its predecessor. The model's ability to ingest direct satellite data also fills critical observation gaps in developing regions where ground rain gauges are sparse.