Google Unveils WeatherNext 3: Most Advanced AI Weather Model Yet

Google has launched WeatherNext 3, its most advanced AI weather forecasting system to date. The model is powered by Google DeepMind and Google Research and offers more accurate forecasts at a higher resolution than its predecessor.

The new model ingests live geostationary satellite data, allowing it to generate fresh forecasts every hour rather than in six-hour blocks. It also trains on sparse weather station readings, enabling it to capture sharp local variations that older models tend to smooth over.

WeatherNext 3 can resolve surface variables like temperature and moisture down to a 5-kilometer grid, other surface variables at 10 kilometers, and atmospheric variables such as wind speed at 25 kilometers. This represents a significant improvement in resolution compared to its predecessor, which was around five times sharper.

The model has been trained on high-quality precipitation datasets, including NASA's satellite-based IMERG product and Google's radar-based global precipitation reanalysis. As a result, it shows meaningful accuracy gains, with medium-range forecasts showing improvements of up to 60% against one satellite-based benchmark.