FrontierAugust 9, 2026via The Decoder
Google Deepmind's WeatherNext predicts cyclone tracks and intensity at the same time
Why it matters
A frontier AI capability milestone (weather prediction) with concrete operational utility and open-source release. Practitioners building with weather models and AI enthusiasts tracking multimodal AI progress both see a genuine benchmark advance.
Key signals
- WeatherNext predicts cyclone tracks and intensity simultaneously
- Outperforms leading operational models by ~1 day lead time
- Matches a decade of traditional weather forecasting progress
- Code and model weights released open-source on GitHub
- Released by Google DeepMind
- Forecasts ~1 day further ahead than leading operational models
- Performance equivalent to ~10 years of traditional weather forecasting progress
- Model weights and code open-sourced on GitHub
- Developed by Google DeepMind
The hook
DeepMind's WeatherNext matches a decade of traditional forecasting progress in one model release—open weights included.
Deepmind's new weather AI forecasts tropical cyclones about a day further ahead than leading operational models, matching a decade of progress in traditional weather forecasting. Code and model weights are open-source on GitHub.