Google launches open source AI for early warnings of cyclones and hurricanes

Researchers from the Google DeepMind and Google Research teams developed and improved the WeatherNext artificial intelligence model, with the aim of offering more effective warnings for cyclones. The creation of the system included the collaboration of important global meteorological agencies, such as the National Hurricane Center, the Cooperative Institute for Research in the Atmosphere and the UK Met Office. The model’s code and parameters were made open source and made available on GitHub, making it easy for other scientists to access and use in their own research.

A detailed study on WeatherNext was published in the renowned journal Nature, with a more accessible version also published on an official Google blog. Forecasting tropical cyclones, which are known as hurricanes or typhoons in different parts of the world, presents a particular challenge. Traditionally, the trajectory of these storms was analyzed by broader global models, while the intensity was calculated by specialized local models, capable of evaluating thermodynamic processes at the center of the cyclone. WeatherNext was trained with nearly 20 terabytes of global atmospheric data and historical information, allowing it to predict both trajectory and intensity in a single model.

Google DeepMind's WeatherNext gives cyclone forecasters an extra day of warning. Published in Nature. Model open-sourced.

→ 3-day forecasts now as accurate as what prior models could do in 2 days
→ Equivalent to a decade of meteorological progress
→ Operates at 28x28km… pic.twitter.com/AC4DDkH49c

— Ramesh Dontha 🦉 (@EntrepreneursAI) August 8, 2026

The speed of the new system is one of its great differences. WeatherNext researchers stated that it is now possible to generate a 15-day forecast in less than a minute using a Tensor Processing Unit (TPU). This agility allows meteorologists to quickly assess the probability distribution of potentially devastating risks, optimizing preparation for and response to extreme events.

The technology giant had already launched the second generation of WeatherNext last year. In addition to work on cyclones, Google research teams have also focused on using artificial intelligence to help predict flash floods, demonstrating an ongoing commitment to applying AI to climate challenges.