Google DeepMind's WeatherNext Gives Forecasters a Full Extra Day to Warn People About Cyclones
The lab is releasing code and model weights for a system that predicted Hurricane Melissa's rapid strengthening and Jamaica landfall during the 2025 hurricane season.
Key points
- Google DeepMind published a paper in Nature showing its WeatherNext AI model predicts cyclone track, intensity and wind structure with roughly 24 extra hours of lead time versus the best prior models.
- During the 2025 hurricane season, WeatherNext helped the US National Hurricane Center forecast Hurricane Melissa's rapid intensification and landfall in Jamaica.
- The model runs on a TPU, a Google-made AI chip, and produces a 15-day forecast in under a minute, generating 1,000 possible storm scenarios at once.
- DeepMind is open-sourcing the code and weights for WeatherNext Cyclones, WeatherNext 2, and a lightweight version called WeatherNext 2-mini that runs in a free public notebook.
- Tropical cyclones have killed more than 700,000 people and caused $1.4 trillion in damage worldwide over the past 50 years.
A hurricane forecast is a race. Every hour a warning goes out earlier is another hour to secure homes or move people off the coast.
Google DeepMind says its AI weather model has just won back a full day of that race.
In a paper published in Nature, the lab reports that WeatherNext predicts where a cyclone will go, how strong it'll get, and how its winds are shaped more accurately than any previous model. Its three-day forecast matches what the best two-day forecasts achieved before, a jump meteorologists equate to roughly a decade of normal progress. DeepMind is open-sourcing the code and model weights so any national weather service or university lab can run them. This is the third AI weather forecasting story we've covered in the past 90 days, and the urgency keeps compounding.
What did the model actually do during a real storm?
It helped forecasters see Hurricane Melissa coming. In 2025, WeatherNext predicted Melissa's rapid intensification and landfall in Jamaica, and the US National Hurricane Center used that signal to issue an unusually early warning. Rapid intensification, when a storm suddenly jumps in strength over a short period, is the scenario forecasters fear most because it leaves the least time to react.
This year the system runs 1,000 simulated versions of each storm at once, a technique called an ensemble forecast. Last year it produced 50 at a time. The jump to 1,000 lets forecasters see not just the most likely path but rare outliers too.
Why is this a big deal technically?
Cyclone forecasting has always been split in two. Where a storm goes is driven by huge global wind patterns, best captured by coarse global models. How strong it gets is driven by churning physics near its eye, which traditionally needs a separate, high-resolution local model.
WeatherNext handles both with one AI. Stranger still, it does the intensity job well while working in fairly chunky 28-by-28-kilometre squares, roughly 100 times coarser than traditional storm models. A smaller version, WeatherNext 2-mini, works at 111-by-111-kilometre resolution and still performs strongly. DeepMind admits it doesn't fully understand why the model is this good at that resolution and calls it an open research question.
The system was trained on nearly 20 terabytes of global atmospheric data plus IBTrACS, an expert-curated record of roughly 5,000 historical storms.
What models are being released, and where can people see them?
| Model | Resolution | What it's for |
|---|---|---|
| WeatherNext Cyclones | 28 km | Cyclone track, intensity, wind structure |
| WeatherNext 2 | 28 km | General global weather, launched October |
| WeatherNext 2-mini | 111 km | Runs on a single TPU in a free Colab notebook |
Forecasts are viewable on Google's Weather Lab site, which now shows temperature, precipitation, wind speed and cyclone tracks together.
What does this mean for people in a storm's path?
Better, earlier warnings from your national weather agency. You won't be downloading WeatherNext yourself. What changes is what the National Hurricane Center, the UK Met Office and other agencies can tell you, and how confidently they can say it days out. The AI is a tool for forecasters, not a replacement for them. For official warnings, keep listening to your local meteorological service.
The honest thing to watch here is operational uptake: an open-source model is only as useful as the agencies willing to integrate it, and that process is slower than a press release.
