Google DeepMind has published peer-reviewed results showing its WeatherNext AI weather model achieves state-of-the-art accuracy in predicting tropical cyclones, gaining an average of a full day or more of predictive lead time over leading operational forecasting systems. The research appeared in the journal Nature on August 6, 2026, and Google is open-sourcing the models alongside the publication.

The breakthrough arrives at a moment of intensifying AI industry coverage, where artificial intelligence is increasingly being applied to high-stakes scientific and public-safety problems. Tropical cyclones have caused more than 700,000 deaths and $1.4 trillion in economic losses over the past 50 years, according to figures cited in the DeepMind announcement, making even modest gains in forecast lead time consequential for evacuation and disaster response.

A Day of Lead Time Compressed Into One System

The headline result is lead time. On cyclones from 2023 through 2025, WeatherNext's track, intensity, and wind-structure predictions carry an average of a day or more of advantage over leading operational models. In practical terms, its three-day forecast matches what prior systems delivered at two days. The researchers describe the improvement as comparable to roughly a decade of operational meteorological progress compressed into a single AI system.

In cyclone forecasting, where evacuation orders, supply staging, and emergency declarations hinge on hours, that margin is the metric that matters most. The paper introduces WeatherNext Cyclones, an AI model that produces ensemble forecasts of a tropical cyclone's track, intensity, and size up to 15 days out.

How the Model Works

The training data combined global atmospheric analysis with the IBTrACS database of nearly 5,000 historical storms. The model uses Functional Generative Networks to produce probability distributions rather than single-point forecasts, and it scales to 1,000-member ensembles — twenty times the 50-member runs of the prior system. Larger ensembles better capture rare but high-consequence events like rapid intensification, the dangerous phenomenon where a storm's wind speeds surge suddenly.

One finding stands out for what it challenges. High spatial resolution has long been considered the price of accurate intensity forecasts. WeatherNext Cyclones runs on 28-by-28-kilometer inputs, roughly 100 times coarser than traditional regional models, and a compact variant called WeatherNext 2-mini operates at 111-by-111 kilometers and still performs well. The paper states that high resolution is not a strict prerequisite for state-of-the-art intensity forecasting, and the researchers flag as an open question how the model extracts intensity signal from coarser data.

Tested During the 2025 Hurricane Season

The evaluation was not purely retrospective. During the 2025 Atlantic hurricane season, the model ran alongside the National Hurricane Center's operational workflow and helped forecasters anticipate Hurricane Melissa's rapid intensification and landfall in Jamaica far enough in advance to support an early warning. The NHC's 2025 verification report documents the season's forecast performance. This year the system is generating 1,000 possible scenarios per cyclone to support forecaster decision-making.

The work was a collaboration between Google DeepMind and Google Research scientists and operational forecasters at the National Hurricane Center, the Cooperative Institute for Research in the Atmosphere (CIRA), and the UK Met Office, several of whom are co-authors on the Nature paper.

Three Open-Source Variants Released

The open-source release covers three model variants:

  • WeatherNext Cyclones — the version that ran during the 2025 hurricane season and whose results appear in the paper
  • WeatherNext 2 — the later update that Google operationalized in October 2025
  • WeatherNext 2-mini — a compact version that runs on a single TPU in a free Google Colab notebook

Code and weights for all three are available on GitHub. A single 15-day forecast takes under a minute on one TPU, meaning the constraint on adoption is now expertise and integration, not compute.

Forecasts are also explorable on Weather Lab, which Google has expanded from cyclone tracking to full global forecasts covering temperature, precipitation, and wind speed. Both the models and Weather Lab sit within Google Earth AI, the company's umbrella effort for applying machine learning to climate and geospatial challenges.

Why It Matters

The result reinforces a broader trend: AI-based weather models are moving from research demonstrations to operational tools faster than many meteorologists expected. By open-sourcing the weights and publishing in a top-tier journal, Google is positioning WeatherNext as a reference system that national meteorological agencies and researchers worldwide can build on.

For communities in cyclone-prone regions, an extra day of warning can be the difference between an orderly evacuation and a chaotic one. As AI weather models continue to mature, the competitive question is shifting from accuracy alone to who can translate forecasts into action fastest.

Stay Ahead of AI

For more on how artificial intelligence is transforming science, infrastructure, and daily life, follow our breaking AI news coverage. We track the research, the companies, and the policies shaping the field in real time.

Read more AI news →