WeatherNext 3 is DeepMind’s new global weather AI model, and it just replaced the forecasting engine behind Google Search, the Gemini app, and Google Maps. No update to install. Billions of people got a better forecast while they slept.
5x sharper, refreshed every hour
The old WeatherNext 2 ran on a 25 km grid, updated every 6 hours. WeatherNext 3 hits 5 km resolution with hourly refreshes — 15-day probabilistic forecasts, 64 ensemble members. The real breakthrough: it’s the first version to ingest live geostationary satellite data, skipping the roughly 6-hour data lag that traditional numerical weather prediction can’t shake. Precipitation accuracy improves up to 50%, with the biggest gains in regions weather agencies have historically underserved.
Built for power grids, not just umbrellas
It also outputs 100-meter wind speed (turbine height), cloud cover, and solar radiation — the numbers grid and renewable operators actually pay for.
The API angle
Developers pull forecast data through the Google Maps Platform Weather API, plus BigQuery, Earth Engine, and Cloud Storage. Obvious use cases: wind farm output prediction, solar yield planning, logistics routing around storms.
You Might Also Like
- Google Ships Gemini api File Search Multimodal rag Page Level Citations and Embedding 2 Baked in
- Gemini Canvas in ai Mode Google Just Turned Search Into a Creative Workspace
- Google Maps ask Maps Turns Every Search Into a Conversation and Kills the old qa Feature
- Google Health Swaps the Fitbit app for a 9 99 Gemini Coach
- Google Search 全面切换 Gemini 3 5 Flash（ai Mode 全球默认 生成式 ui） 2 Billion Users and Blue li
Top AI Product 
Leave a comment