Google says its new AI weather system can talk to satellites and deliver clear forecasts every hour

Google has launched WeatherNext 3, an AI weather model that can deliver forecasts every hour at up to 5km resolution using real-time satellite data. The model also promises better rain forecasts and is coming to Google Search, Gemini, Maps and other Google products.Google launches WeatherNext 3 AI weather model. (image credit: Google)Google has introduced WeatherNext 3, its latest AI model for weather forecasting, which can generate high-resolution forecasts every hour using data from geostationary satellites.Google DeepMind and Google Research describe WeatherNext 3 as “the most advanced and accurate global weather model to date”, based on independent live evaluations by Brightband. Unlike earlier AI weather models that mainly learned from forecasts generated by traditional systems, the new model can learn directly from real-time weather observations.Google says WeatherNext 3 will also power weather information across Google Search, Gemini, Google Maps, Google Maps Platform and Google Earth Engine.WeatherNext 3 can forecast weather at 5km resolutionOne of the biggest upgrades is the level of detail WeatherNext 3 can provide. Google says the model can forecast key surface variables, including temperature and moisture, at a resolution of 5 kilometres. Other surface variables are available at 10km resolution, while atmospheric variables such as wind speed can be forecast at 25km. Overall, Google says this provides “a global weather picture roughly five times sharper than our previous model, WeatherNext 2”, which produced forecasts on a 25km grid in six-hour increments.WeatherNext 3 can refresh forecasts every hour, giving it a better chance of tracking fast-changing conditions such as storms, fronts and precipitation systems. Google says the model's biggest improvement comes from the data it uses. Instead of depending mainly on numerical weather prediction models that can have a six-hour data lag, WeatherNext 3 uses a mosaic of live, global geostationary satellite data.Google says this gives the model “a rich, continuously updating view of the atmosphere”, allowing it to produce a new forecast every hour based on the latest satellite observations.The model also uses weather station observations to account for local factors such as mountains, valleys and coastlines. Google says this is particularly important for regions across Latin America, Africa and Asia-Pacific, which have historically had less access to high-resolution forecasting because of the cost of traditional supercomputing-based models.Google says rain forecasts are also getting betterGoogle says it trained the model using precipitation data from NASA's Integrated Multi-satellite Retrievals for GPM (IMERG) and its own global precipitation reanalysis.The company says evaluations showed a Continuous Ranked Probability Score improvement of up to 60 per cent against IMERG, 30 per cent against MRMS and 10 per cent against rain gauge measurements for early lead times.Google has also designed WeatherNext 3 for renewable energy planning. The model can forecast 100-metre wind speeds, roughly the height of wind turbines, along with high-resolution cloud cover and solar radiation. This could help wind and solar operators estimate how much electricity their projects may generate.WeatherNext 3 is already being integrated into Google's products. The company says that when planning a day or more ahead, users will see “up to 50% more accurate precipitation forecasts”, with the biggest improvements expected in regions where forecasts have historically been less reliable.- EndsPublished On: Sep 4, 2026 11:37 IST