Tech news in 3 minutes
Google’s latest AI weather model gives you no excuse to forget your umbrella
Google DeepMind and Google Research have released WeatherNext 3, a new AI weather forecasting model that delivers higher accuracy, finer resolution, and hourly predictions, and will soon power weather information in Google Search, Maps, and Gemini. The model, announced today, achieved the top score on Brightband’s OperationalWeatherBench benchmark, outperforming AI models from Google, Microsoft, Nvidia, and the European Centre for Medium-Range Weather Forecasts (ECMWF), as well as traditional forecasts from the U.S. National Weather Service and ECMWF. WeatherNext 3 predicts key variables like temperature, wind speed, and humidity at a resolution of 5 km—down from the typical 15–25 km of other AI models—and produces hourly forecasts instead of the standard six-hour intervals. Its rain prediction accuracy is 60% improved over WeatherNext 2. The model has 2.4 times more parameters than its predecessor and is the first AI model, according to Google, to directly incorporate raw satellite observations for a high-resolution global forecast (though AI startup WindBorne says its WeatherMesh 6 has done so since late 2025). Google senior staff engineer Samier Merchant told TechCrunch this is the first time core weather variables will feed Google products. DeepMind researcher Ferran Alet noted that machine learning targets approximate noisy physics from incomplete data, and that higher-resolution forecasts of wind, rain, and cloud cover can make renewable energy projects more dependable. Brightband atmospheric scientist Daniel Rothenberg highlighted the model’s ability to predict what a specific weather station—like Denver’s airport—will measure hourly, connecting forecasting closer to ground truth. Bill Gates recently cited AI-powered weather forecasting as crucial for improving crop yields in developing countries. WeatherNext 3 will be available to users and researchers on Google Cloud platforms.