Skip to content

News with true faith

Top News

Google Unveils WeatherNext 3 AI Weather Forecast Model

Google DeepMind's WeatherNext 3 uses live satellite data to forecast weather at 5-kilometer resolution, improving rain prediction by up to 50%.

Google Unveils WeatherNext 3 AI Weather Forecast Model

Google DeepMind has announced WeatherNext 3, a new version of its artificial intelligence weather forecasting model, aiming to deliver more detailed and higher-resolution predictions of core weather variables such as temperature, humidity and wind speed.

Google’ın yeni yapay zekası hava durumu tahmininde iddialı
Google DeepMind has announced WeatherNext 3, a new version of its AI-powered weather forecasting model, aiming to deliver more detailed and higher-resolution forecasts of core weather variables such as temperature, humidity and wind speed.

One of the biggest differences from the previous generation is forecast resolution. WeatherNext 2 used a grid of 25-kilometer squares, while the new model can produce forecasts over 5-kilometer squares. Forecasts can also be delivered hourly.

Live satellite data added to the model

According to DeepMind, the added detail in WeatherNext 3 is built on new data sources. Data drawn from numerical weather prediction models used in earlier versions carried a delay of roughly six hours. WeatherNext 3 instead feeds live satellite data directly into the model.

Google’ın yeni yapay zekası hava durumu tahmininde iddialı
The approach allows the constantly updated state of the atmosphere to be fed into the model, which also draws on sparser meteorological station data, enabling more detailed forecasts of variables such as humidity that can vary sharply over short distances.

This approach allows the constantly updated state of the atmosphere to be fed into the model. The system also draws on data from meteorological stations, which are more sparsely distributed. DeepMind said this combination allows for more detailed forecasts of weather data such as humidity, which can change significantly even over short distances.

Rain forecasts improved by up to 50%

One of the standout improvements in WeatherNext 3 is rainfall forecasting. Google said it combined precipitation data from NASA with its own satellite analysis to improve the model's ability to predict types of precipitation such as rain and snow.

DeepMind said the new model can achieve up to 50% greater accuracy in precipitation forecasts. The company said the improvement is especially pronounced in regions where weather forecasts have historically been less reliable.

Where the model will be used

Google said the high-resolution forecasting capabilities of WeatherNext 3 could be useful in particular for the renewable energy sector. The model will also power Google Search, the Gemini app and Google Maps, along with the Google Maps Weather API and Google Earth Engine platforms.

Users who want to try the model directly can do so through Weather Lab. Google also made its first WeatherNext model open source as of August 2026.

Related

Leave a comment

Your email address will not be published. Required fields are marked *