Weather forecasts are about to become more personal, more detailed, and more tightly connected to the apps people use every day. On September 3, 2026, scientists at Google DeepMind and Google Research released WeatherNext 3, an artificial intelligence model built to push weather prediction toward higher resolution and faster updates.
At the same time, the World Meteorological Organization warned that El Niño is set to intensify, creating risks of more extreme weather through the coming months and into 2027. Better forecasts are arriving as the need for earlier warnings grows.
A New Weather Engine For Google’s Products
Google says WeatherNext 3 will begin feeding weather information into Google Search, Google Maps, and Gemini, while also becoming available through Google’s cloud platforms. That brings the model closer to the tools people already use to plan travel, find information, and make decisions.
“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.
WeatherNext 3 has already proven to be the most accurate model among leading contenders tested on Operational WeatherBench. Its results beat other deep-learning models built by Google, Microsoft, NVIDIA, and the European Center for Medium-Range Weather Forecasting, along with traditional forecasts from the US National Weather Service and the European Center for Medium-Range Weather Forecasting.
That contest reflects a major shift in meteorology. Most weather forecasts still come from government-owned supercomputers, which are expensive and comparatively slow. After the European Center for Medium-Range Weather Forecasting released weather data from government supercomputers in June 2018, deep-learning researchers began training models that could produce predictions faster while maintaining comparable accuracy.
More Detail, More Updates, Fewer Gaps
WeatherNext 3 can predict key variables down to a resolution of 5km, giving the model a more granular view of weather conditions. Its evaluations on rain improved by 60% over WeatherNext 2, and it can produce hourly forecasts instead of predictions every six hours.
The model is also larger, with 2.4 times more parameters than its predecessor. Google trained it to target forecasts to specific weather data stations, a design that aims to connect broad global prediction with conditions measured at individual locations.
“Adding a capability where this model is now also predicting, say, what Denver’s airport’s weather station is going to measure on an hourly basis, just connects that forecasting task closer to the core,” said Daniel Rothenberg, an atmospheric scientist at Brightband.
WeatherNext 3 can ingest weather satellite data collected in real time on an hourly basis. Google states that it is the “first” AI model to directly incorporate raw observations for a high-resolution global forecast, though WindBorne’s WeatherMesh 6 has incorporated raw observations since late 2025.
Both WeatherNext 3 and WeatherMesh 6 still rely on national weather datasets for forecasts. The European and US weather agencies are already using AI models in their forecast products, showing how machine learning is moving into both commercial tools and official weather services.
Ferran Alet, a staff research scientist manager at DeepMind, described the challenge behind these systems: “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.”
Why The Timing Matters
The forecast advances arrive as the World Meteorological Organization warns about an El Niño event with global consequences. The WMO said the ocean warming climate phenomenon will build into “a very strong event,” bringing big impacts on rainfall and temperature patterns.
Forecasts show a near 100% likelihood of El Niño continuing through February 2027. The event is associated with “exceptionally warm tropical Pacific Ocean conditions,” and its impacts are expected to persist well into 2027.
The WMO called for unprecedented action by the U.N. and national agencies, including stronger early-warning systems. Celeste Saulo, the WMO Secretary-General, said, “Never before in the 50-year history of the World Meteorological Organization have we launched such a major mobilization with National Meteorological and Hydrological Services who are on the frontline of delivering the forecasts and services to save lives and livelihoods.”
Saulo said the WMO is working with partners to provide climate intelligence and insights for disaster management and climate-sensitive sectors. She also stated that impacts from droughts and floods are expected to increase as El Niño intensifies.
The warning carries a human cost already visible in the Himalayan glacial floods in Nepal, where the Nepalese Army became involved in flood rescue efforts and reported that the death toll topped 1,000 on September 3, 2026.
U.N. Secretary-General António Guterres put the broader danger in direct terms: “The science leaves no room for doubt: the planet is in uncharted waters, and those waters are heating up. Sea surface temperatures are rising, temperatures keep climbing, and the world is in the danger zone of extreme weather. The race now is between rising risks and our commitment to take climate action and protect people. We must win that race.”
WeatherNext 3 does not remove the uncertainty from weather, but it brings faster updates, finer detail, and a direct path into everyday technology. As El Niño raises the stakes through 2027, the value of turning raw observations into timely warnings will only grow.
Based on
