Google DeepMind has released WeatherNext 3, an AI weather model that pulls in live satellite data to produce hourly forecasts on a 5-kilometer grid, a fivefold jump in detail over its predecessor. The new system is also designed to predict rain and snow more accurately, with the company claiming up to 50 percent better precipitation forecasts in regions where traditional predictions have struggled.
The shift from older numerical weather prediction methods to a model trained on continuously updating satellite observations is the core change. That richer view of the atmosphere lets WeatherNext 3 capture fast-changing variables like humidity, which can shift dramatically from one kilometer to the next.
What WeatherNext 3 does differently
WeatherNext 2 ran on a 25-kilometer square grid. WeatherNext 3 shrinks that to a 5-kilometer square grid and adds an hourly cadence, giving users a much sharper look at near-term conditions. DeepMind says the model also excels at niche variables that matter to specific industries, including 100-meter wind speeds (roughly at turbine height) for wind-energy output and high-resolution cloud cover plus sun radiation data for solar farms.
Rather than depending on numerical weather prediction models, which carry a roughly six-hour data lag, WeatherNext 3 ingests live satellite data alongside sparse weather station readings. The result is what DeepMind describes as a "rich, continuously updating view of the atmosphere."
Why precipitation forecasting improves
The model combines precipitation data from NASA with Google’s own satellite analysis. DeepMind reports that WeatherNext 3 produces "up to 50 percent more accurate precipitation forecasts, with the greatest improvements in regions where forecasts have historically been less reliable." That matters for users in regions where existing weather services have a track record of getting rain and snow wrong.
Where the forecasts will appear
WeatherNext 3 will power weather experiences in Google Search, the Gemini app, and Google Maps. It will also be available through the Google Maps Weather API and Google Earth Engine, giving developers a path to integrate the forecasts into their own products.
How to try the model
Anyone curious to see the forecasts in action can experiment with the model now in the Google Weather Lab. The earlier WeatherNext model is open-source as of August 2026, so developers looking to build on top of the technology have a starting point.
What it means for renewable energy operators
For renewable energy companies, the new wind and solar variables could translate into more precise energy output forecasts. Wind farms depend on knowing exactly how fast air is moving at hub height, since power output scales sharply with wind speed. Solar farms need accurate cloud cover and radiation readings to predict generation. WeatherNext 3’s higher resolution and hourly cadence target both needs directly.
What it means for everyday users
For the average person checking the forecast, the practical effect is more reliable short-term predictions, especially for precipitation in regions that have historically been underserved by weather services. The hourly cadence on a 5-kilometer grid means a local forecast that better reflects conditions in a specific neighborhood rather than averaging across a much wider area.
FAQ
What is WeatherNext 3?
WeatherNext 3 is the latest AI weather model from Google DeepMind. It uses live satellite data to produce hourly forecasts at a 5-kilometer resolution, up from the 25-kilometer grid of its predecessor.
How much better are the precipitation forecasts?
DeepMind says WeatherNext 3 produces up to 50 percent more accurate precipitation forecasts, with the largest gains in regions where forecasts have historically been less reliable.
Where can I access WeatherNext 3 forecasts?
The model powers weather features in Google Search, the Gemini app, and Google Maps. Developers can use the Google Maps Weather API or Google Earth Engine, and curious users can experiment with the model directly in the Google Weather Lab.
Related coverage
- Google TimesFM-3 Forecasts Sales From Weather, Discounts, and Related Data
- Ox Alpha: A Free Million-Token AI Model Now Live on OpenRouter
This article summarizes reporting from engadget.com. See our editorial disclaimer for how our articles are produced.
Run a free scan to see your AI Visibility Score, SEO rating, and local citation accuracy.