Google DeepMind has open-sourced WeatherNext, an AI weather model that predicted hurricane tracks one day earlier than traditional forecasting systems, according to research published in Nature.
The model matched the accuracy of conventional 2-day forecasts using only a 3-day outlook, making it a potential breakthrough for faster hurricane warnings.
What happened
Google DeepMind has released the code and weights for WeatherNext, an AI weather model, after publishing research in Nature showing it predicted hurricane tracks one day earlier than traditional forecasting systems.
Why it matters
The model matches the accuracy of conventional forecasts at the 2-day horizon using only a 3-day outlook, suggesting AI-driven weather prediction could deliver faster, more actionable hurricane warnings for emergency preparedness and public safety.
What to watch
The open-source release means researchers and weather services worldwide can now test and integrate WeatherNext into their own forecasting workflows, potentially accelerating adoption of AI in meteorology.
Google DeepMind announced the open-source release of WeatherNext, an AI-based weather forecasting model, following the publication of research in Nature that documents its hurricane prediction performance. In tests, WeatherNext predicted hurricane tracks one day earlier than traditional forecasting systems. The model's accuracy at the 3-day forecast window matched what conventional systems achieve at the 2-day mark, effectively compressing forecast lead time by one full day. This capability could be transformative for hurricane preparedness: a one-day advance in reliable track prediction gives meteorologists, emergency managers, and the public additional time to issue warnings, plan evacuations, and mobilize resources. By releasing both the code and model weights as open-source software, Google DeepMind has made the technology available to the broader research and operational weather forecasting community, reducing barriers to adoption and enabling weather services worldwide to test, validate, and potentially integrate WeatherNext into their own systems.
Weather forecasting has traditionally relied on physics-based computational models that simulate the atmosphere with incrementally improving precision. Google DeepMind's WeatherNext represents a shift toward AI-driven prediction: by matching conventional 2-day accuracy at a 3-day horizon, it compresses the timeline in which forecasters can issue reliable warnings. The one-day advance in hurricane track prediction is particularly significant for tropical cyclones, where track uncertainty compounds rapidly and early, accurate warnings are essential for evacuation and preparation. By open-sourcing the model, Google DeepMind is democratizing access to this capability—a move that signals confidence in the approach and may encourage other organizations to build on or integrate the model into their own forecasting infrastructure.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
CloudSEK identified more than 2,500 organisations potentially exposed by a March 2026 LiteLLM incident, an ope…
At the Ai4 conference in Las Vegas, Nobel Prize winner Geoffrey Hinton, World Labs CEO Fei-Fei Li, and Courser…
Nvidia is developing a trillion-parameter model in its Nemotron 4 family as part of an expanded open-source st…
IBM and Together AI announced a collaboration to deploy a dedicated Nvidia-powered inference cluster on IBM Cl…
Meta CEO Mark Zuckerberg published a letter Monday titled "The Future is for Everyone," stating superintellige…
A new platform called frontier.fast has launched an open competition where anyone can submit code patches to m…
The AI news that matters, in one minute each morning.
Sign up free