Researcher at Google DeepMind and Google Research have unveiled a new artificial intelligence model for weather forecasting that offers clearer insights into our changing atmosphere and delivers more frequent predictions than ever before. As of 2026, this cutting-edge model, named WeatherNext 3, represents the latest breakthrough in the deep learning revolution reshaping meteorology.
Google plans to integrate WeatherNext 3 into the weather information displayed across its ecosystem, including Search, Google Maps, and Gemini. The model will also be accessible to users and researchers through Google’s cloud platforms, democratizing access to state-of-the-art forecasting capabilities.
“This is going to be the first time that some of the core variables feed and power a lot of the Google products,” said Samier Merchant, a Google senior staff engineer, in an interview with TechCrunch.
Proven Accuracy Among Leading Models
WeatherNext 3 has already established itself as the most accurate model among top contenders tested on Operational WeatherBench, a benchmarking utility developed by startup Brightband to compare AI forecasts. The evaluation assesses metrics such as temperature, windspeed, and humidity. Notably, it outperforms not only other deep learning models from Google, Microsoft, Nvidia, and the European Centre for Medium-Range Weather Forecasts (ECMWF) but also surpasses traditional forecasts from the U.S. National Weather Service and ECMWF, marking a significant milestone for AI-driven meteorology.
The Shift from Supercomputers to Deep Learning
Traditional weather forecasting relies on government-owned supercomputers that laboriously process vast sets of mathematical equations describing atmospheric physics. While remarkably accurate, these systems are costly and relatively slow. In 2018, the ECMWF released more than half a century of such data, enabling deep learning researchers to train models that make predictions with comparable accuracy but at a fraction of the time and cost.
“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, and so it learns patterns from a lot of data,” explained Ferran Alet, a staff research scientist manager at DeepMind.
Addressing Key Challenges in AI Forecasting
Since the initial deep learning breakthroughs, model developers have focused on overcoming persistent limitations in AI forecasting. Earlier models tended to predict over broad areas of 15 to 25 square kilometers—too coarse for hyperlocal use—and often struggled with precipitation accuracy. Furthermore, they depended heavily on pre-formatted datasets.
WeatherNext 3 addresses these challenges by improving resolution and rainfall prediction, marking a crucial step toward more reliable, actionable forecasts. As the model rolls out across Google’s platforms, users around the world will benefit from enhanced, AI-powered weather insights, making it easier than ever to know when to grab an umbrella.
via TechCrunch AI