Google has released WeatherNext 3, an updated artificial intelligence model for weather forecasting from Google DeepMind and Google Research, according to TechCrunch and The Verge. The company is putting the model into consumer-facing products including Search, Google Maps, and Gemini, while also making it available through Google’s cloud platforms, TechCrunch reports. The core change is granularity. Both outlets report that WeatherNext 3 can produce forecasts every hour and visualize some weather variables at up to 5-kilometer resolution. The Verge says the previous WeatherNext 2 model produced forecasts every six hours on a 25-kilometer grid. Google is positioning the model as a step forward for short-term and localized forecasts, particularly precipitation. The Verge reports that WeatherNext 3 uses real-time satellite observations and “fresher and richer” observational data, according to Google Research engineer Samier Merchant. Google says that helps the model produce a sharper global picture and improve forecasts for rain and snowfall, especially in fast-moving weather systems. The precipitation claims should be read with their stated scopes. TechCrunch reports that WeatherNext 3’s evaluations on rain are 60% improved over WeatherNext 2. The Verge separately reports that Google says users will see precipitation forecasts that are up to 50% more accurate when looking at least a day ahead. Those figures are not directly interchangeable, but both point to precipitation as a main target of the update. The model also reflects a broader shift in meteorology. Traditional forecasting relies heavily on government supercomputers that simulate atmospheric physics through complex equations. The Verge and TechCrunch both note that AI models instead learn patterns from weather data, enabling faster predictions than conventional numerical systems, though they still depend on observational and historical data pipelines. TechCrunch reports that WeatherNext 3 has already ranked as the most accurate among leading systems tested on Operational WeatherBench, a comparison tool built by Brightband. According to TechCrunch, those tests cover variables including temperature, wind speed, and humidity, and WeatherNext 3 outperformed other deep-learning models from Google, Microsoft, Nvidia, and the European Centre for Medium-Range Weather Forecasts, as well as traditional forecasts from the U.S. National Weather Service and the ECMWF. The Verge does not report that benchmark result, so it remains a single-source claim within this cluster. Google is also tying the model to energy forecasting. The Verge reports that WeatherNext 3 was designed to produce forecasts for renewable-energy generation, including wind speed at 100 meters, roughly turbine height. Ferran Alet, a research scientist at Google DeepMind, told The Verge that improving renewable-energy forecasting matters as energy needs rise. For users, the immediate change is that Google’s weather surfaces should become more tightly coupled to its latest AI forecast model. For researchers and developers, the cloud-platform availability matters because it can expose WeatherNext 3 beyond Google’s own apps. The evidence here supports the product rollout and the broad technical improvements; the strongest accuracy claims still require careful attribution to Google or to TechCrunch’s reporting on Operational WeatherBench. Who benefits: Google users may see more detailed forecasts in core products, particularly for precipitation. Developers, researchers, and weather-sensitive operators could benefit if Google’s cloud access makes the model usable outside Google’s own apps. Who's exposed: It is too early to tell what competitive effect WeatherNext 3 will have on traditional forecast providers or rival AI weather-model developers. TechCrunch reports it ranked as the most accurate among leading systems tested on Operational WeatherBench.