Google DeepMind Unveils WeatherNext 3: Hourly AI Weather Forecasts
Google DeepMind launched WeatherNext 3, its most advanced global weather AI: hourly forecasts, 5 km resolution and direct satellite data, feeding into Search, Maps and Gemini.
What happened: DeepMind's most advanced weather AI yet
Google DeepMind has introduced WeatherNext 3, which the company calls its most advanced global weather model. The key breakthrough: it is the first global weather model that generates forecasts every hour of the day, rather than a few times per day.
Hourly data straight from satellites
Unlike earlier systems, WeatherNext 3 is an ensemble model that draws directly on raw satellite imagery. That lets it track fast-changing conditions such as rain and snow in near real time. It predicts weather-station-targeted temperature and humidity at 5 km resolution, and other surface variables such as wind at 10 km — with significantly higher accuracy even in locations the model never saw during training.
Built for industry: from Search to wind farms
The model is designed for enterprise use as well: for wind and solar farm operators it covers variables such as radiation and cloud cover. Forecasts are being integrated into Google Search, Maps and Gemini, while enterprise access is available through BigQuery, Earth Engine, Google Maps Platform and Google Cloud Storage. According to DeepMind, WeatherNext helped the US National Hurricane Center better predict Hurricane Melissa's historic landfall in Jamaica.