Google DeepMind Launches WeatherNext 3, an AI Weather Model That Forecasts Hourly at 5-Kilometer Resolution

Overview

Google DeepMind and Google Research today introduced WeatherNext 3, describing it as “the most advanced and accurate global weather model to date, according to independent live evaluations by Brightband,” according to Google. The model begins rolling out today inside Google Search, the Gemini app, Google Maps, the Google Maps Platform Weather API, and Google Earth Engine, according to Google.

What We Know

WeatherNext 3 forecasts key surface variables such as temperature and moisture at 5-kilometer resolution, other surface variables at 10 kilometers, and atmospheric variables like wind speed at 25 kilometers, according to Google. That is “roughly five times sharper” than its predecessor, WeatherNext 2, which produced forecasts on a 25-kilometer grid in 6-hour increments, according to Google.

The new model generates a fresh forecast every hour instead of every six hours by ingesting live geostationary satellite data, rather than relying solely on the numerical weather prediction data — which carries a six-hour data lag — that most AI weather models, including WeatherNext 2, are trained on, according to Google. WeatherNext 3 is also “a larger model, with 2.4 times more parameters than its predecessor,” and its designers tailored the targets for its decoder heads, according to TechCrunch.

On precipitation, Google reports a Continuous Ranked Probability Score improvement “of up to 60% against IMERG, 30% for MRMS, and 10% against rain gauge measurements for early lead times,” according to Google. The model trains in part on NASA’s Integrated Multi-satellite Retrievals for GPM (IMERG) precipitation data and Google’s own satellite-radar-based precipitation reanalysis, according to Google. TechCrunch separately reported that the model’s “evaluations on rain are 60% improved over WeatherNext 2,” according to TechCrunch. For longer-range planning, Google says people will see “up to 50% more accurate precipitation forecasts” when planning a day or more ahead, with the largest gains in regions where forecasts have historically been less reliable, according to Google.

WeatherNext 3 also forecasts 100-meter wind speeds — roughly turbine height — along with cloud cover and solar radiation levels, which Google says can help grid operators and renewable-energy developers plan generation, according to Google. Google says the added resolution is especially significant for Latin America, Africa, and Asia-Pacific, regions it says have “historically been underserved by high-resolution forecasting due to the immense supercomputing costs of traditional regional models,” according to Google.

The model was newly trained to target forecasts to specific weather stations, which Google says produces more granular predictions and lets researchers evaluate results against ground-truth data, according to Google. Daniel Rothenberg, an atmospheric scientist at Brightband, said “the idea, with a lot of AI applications, is to try to run tasks as end-to-end as possible.” He added that “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,” according to TechCrunch.

Google says the new model outperforms other AI weather models tested on Operational WeatherBench, an evaluation tool built by the startup Brightband, and also beats out weather models built by Microsoft, Nvidia, and the European Center for Medium-Range Weather Forecasting, as well as traditional forecasts from the US National Weather Service and the ECMWF, according to TechCrunch. Samier Merchant, a Google senior staff engineer, told TechCrunch: “This is going to be the first time that some of the core variables feed and power a lot of the Google products,” according to TechCrunch.

Ferran Alet, a staff research scientist manager at DeepMind, described the underlying challenge: “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,” according to TechCrunch. Alet added: “At the end of the day, I think Google is about providing useful information to the user, and a lot of what users are looking for has to do with the weather in some way or another,” according to TechCrunch.

Beyond the consumer apps, Google says the underlying forecast data is available for researchers, developers, and businesses to query in BigQuery and Earth Engine, or to bulk-download from Google Cloud Storage, according to Google.

What We Don’t Know

Google describes WeatherNext 3 as the “first” AI model to directly incorporate raw satellite observations for a high-resolution global forecast, but the competing AI weather startup WindBorne says its own model, WeatherMesh 6, “has been incorporating raw observations from its fleet of weather balloons and other sources since late 2025.” Asked about the discrepancy, Google “pointed out that its forecasts are higher resolution across the globe,” according to TechCrunch. TechCrunch notes that “both models still rely on national weather datasets to perform forecasts, so more work will be required for true direct data assimilation,” according to TechCrunch.

Google’s own account of the 60% precipitation-forecasting gain is framed as a Continuous Ranked Probability Score improvement measured against the IMERG satellite dataset, while TechCrunch describes the same figure as an improvement “over WeatherNext 2.” Neither source clarifies whether these are two descriptions of the same evaluation or separate comparisons against different baselines.