Google's WeatherNext 3 Sees the Sky Better Than the Weather Service

Google's new AI weather model beats government supercomputers on accuracy — and now it's going straight into Search, Maps, and Gemini.

Published: 2026-09-04 Category: Quick Take Sources: TechCrunch

The Story

Scientists at Google DeepMind and Google Research released WeatherNext 3, a new AI model for weather forecasting that the company says sees the atmosphere more clearly and predicts its behavior more often. It's the latest wave of the deep-learning sea change in meteorology, and Google says the model will start feeding into the weather info users see in Search, Google Maps, and Gemini, as well as being available to users and researchers on Google's cloud platforms.

"This is going to be the first time that some of the core variables feed and power a lot of the Google products," Samier Merchant, a Google senior staff engineer, told TechCrunch.

The new model has already proven the most accurate among leading contenders on Operational WeatherBench, a utility for comparing AI forecasts built by startup Brightband, which measures metrics like temperature, windspeed, and humidity. It not only beats other deep-learning models from Google, Microsoft, Nvidia, and the European Center for Medium-Range Weather Forecasting (ECMWF), but also beats traditional forecasts from the U.S. National Weather Service and the ECMWF.

Why It Matters

Most weather forecasts still come from government-owned supercomputers churning through physics equations — systems that are remarkably accurate but expensive and comparatively slow. When ECMWF released more than half a century of that data in 2018, deep-learning researchers began training models that predict far faster with comparable accuracy.

Since then, model-makers have attacked the key weaknesses of AI forecasting: they tend to forecast over a wider area (15-25 square km) than is truly useful, they're not always great with rain, and they still depend on government-formatted datasets. WeatherNext 3 takes on all three challenges. On key variables it can predict down to a resolution of 5 km, its rain evaluations are 60% improved over WeatherNext 2, and it can now produce hourly forecasts instead of the standard every-six-hours prediction.

Ferran Alet, a staff research scientist manager at DeepMind, framed it plainly: "Weather is chaotic, and so small differences really start to perturb massively. Machine learning targets the problem we are really solving, which is approximate noisy physics from incomplete information and finite compute."

The Takeaway

The accuracy numbers matter, but the distribution is the real story. Weather is one of the highest-trust, most universally consumed information products on the planet, and Google is now injecting an AI model directly into Search, Maps, and Gemini — the surfaces where billions of people check the forecast daily. That's a quiet but enormous deployment of AI into everyday decision-making, and it will be a live, global stress test of how trust in AI forecasts holds up against government-issued ones.

Source: TechCrunch, "Google's latest AI weather model gives you no excuse to forget your umbrella" (Sept 3, 2026).