DeepMind Open-Sources WeatherNext, Boosting Cyclone Forecast Accuracy by a Decade
GoogleDeepMind · x · 2026-08-06
Google DeepMind announced the open-sourcing of its WeatherNext AI weather model's code and weights, alongside a new paper published in Nature.
The model achieves state-of-the-art accuracy in predicting cyclone tracks, intensity, and wind structure. On average, its 3-day forecasts are as accurate as previous 2-day forecasts, giving forecasters an extra day of lead time—roughly equivalent to a decade’s worth of meteorological progress.
- Real-world impact: During the 2025 hurricane season, the model accurately predicted Hurricane Melissa's rapid intensification and Category 5 landfall 5 days in advance with 80% confidence.
- Technical details: Trained on global atmospheric data and nearly 5,000 historical cyclones, it generates a 15-day probabilistic forecast scenario in under a minute on a TPU.
- Future plans: WeatherLab will now provide 1,000 probabilistic predictions per storm to support forecasters.
Related event: Google Open-Sources WeatherNext AI for Cyclone Forecasting(9 posts)→
More from Research
- Free Open-Source Visual Masterclass: 12 Chapters on LLMs — mdancho84 · 2026-08-07
- Stanford's Evo 2 AI Tool Designs Novel Bacteriophage to Kill E. coli — HumbleRestaurant790 · 2026-08-07
- Quanta Deep Dive: Why AI is Cracking the Legendary Erdős Math Problems — theomitsa · 2026-08-07
- Scale AI Paper Proposes Interaction-Centric Taxonomy to Localize Agent Failures — theomitsa · 2026-08-07
- Can Biological AI Achieve Robust Scaling Laws Like Traditional ML? — josephdviviano · 2026-08-07
- NVIDIA's Cross-Model KV Cache Transfer Speeds Up Inference by 25x — theomitsa · 2026-08-07