WeatherNext
Google DeepMind's family of AI weather models, including a 2026 version that forecasts tropical cyclone tracks and strength in one system.
WeatherNext is Google DeepMind’s line of weather forecasting models. Instead of simulating the physics of the atmosphere the way traditional forecasting does, these models learn patterns from decades of past weather data and then predict what comes next. That makes them dramatically faster: a 15-day forecast runs in under a minute on one of Google’s AI chips, so the system can play out a thousand possible futures for the same storm.
The 2026 addition, WeatherNext Cyclones, tackles tropical storms and does something the older systems could not: predict both where a storm will go and how strong it will get, in a single model. Published in Nature and built with the US National Hurricane Center and the UK Met Office, it sees roughly one day further ahead than the best operational systems, which is about what conventional forecasting gained in the previous decade. Code and weights are public on GitHub, and DeepMind stresses it supports forecasters rather than replacing them.