Google's New AI Weather Model Gives You Rain Forecasts Five Times Sharper Than Before

Key points

• Google launched WeatherNext 3 in 2025, an AI weather model that updates forecasts every hour using live satellite data.
• The new model produces weather maps at up to a 5-kilometre grid, five times sharper than its predecessor, WeatherNext 2, which worked on a 25-kilometre grid.
• Precipitation forecasts improve by up to 50 percent for events at least one day away, Google says.
• WeatherNext 3 is now built into Google Search, Maps, and Gemini.
• The model also predicts wind speeds at turbine height to help plan renewable energy output.

When a fast-moving storm rolls in, the difference between a six-hour-old forecast and a one-hour-old forecast can be the difference between grabbing an umbrella and getting soaked. Google's new WeatherNext 3 model, announced this week, aims to close that gap.

What actually changed?

WeatherNext 3 pulls in fresh satellite images every hour and builds a new forecast each time. Its predecessor, WeatherNext 2, updated every six hours on a 25-kilometre grid. The new model reads temperature and moisture at up to a 5-kilometre resolution, meaning it can pick out detail roughly the size of a small town rather than a large county.

Traditional weather forecasting runs on supercomputers that solve the physics equations of the atmosphere. Accurate, but slow. AI models, by contrast, learn patterns from huge libraries of historical weather data and can produce a prediction in a fraction of the time. WeatherNext 3 adds live satellite feeds on top of that approach, so it is not just working from history.

"One of the main developments is for WeatherNext 3 to go beyond what data most global AI models train on," Samier Merchant, a research engineer at Google Research, told The Verge. "We're able to use fresher and richer observational data sets."

Who benefits most from this?

People in regions with few ground-based rain gauges benefit most. Much of Asia, Africa, and Latin America falls into this category, and satellite data fills the gap those missing gauges leave.

Google says the improvement in precipitation accuracy reaches up to 50 percent for forecasts looking at least a day ahead. For fast-moving weather systems, including sudden heavy rain and snow, that sharper picture matters most.

The model also generates wind-speed predictions at roughly 100 metres off the ground, about the height of a wind turbine. That detail helps energy companies plan how much power a wind farm will produce on a given day. Ferran Alet, a research scientist at Google DeepMind, told The Verge that making renewable energy more reliable is a key motivation, partly because AI data centres themselves need a lot of power.

Will this replace your regular weather app?

Not exactly. WeatherNext 3 is woven into Google Search, Maps, and the Gemini AI assistant, so you are already using it if you ask Google about tomorrow's weather. But it sits alongside traditional physics-based models rather than replacing them. Weather agencies still look at several forecasts before issuing warnings, and WeatherNext 3 itself trains partly on data produced by those older models.

Google has also been working with the US National Hurricane Center and agencies in Asia to put the technology to practical use.

Common questions

Does this change the weather info I see in Google Search?

Yes. WeatherNext 3 is already live inside Google Search, Maps, and Gemini, so any forecast you pull up through those products now draws on the new model.

Is this available as a separate app?

No standalone app exists. The improvements show up automatically inside existing Google products at no extra cost.

Is an AI weather model trustworthy on its own?

It is one useful input, not the whole picture. Meteorologists and weather agencies combine AI forecasts with traditional physics-based models before issuing official warnings.