Google DeepMind’s WeatherNext 3 Trains on Weather Station Observations to Deliver 5 km Global Forecasts, Refreshed Every Hour

- Google DeepMind's WeatherNext 3 just made classic weather models look like a DeFi whitepaper — confident, six hours late, and wrong about rain. Hourly initialization from live satellites, ~5 km resolution, 64-member ensembles out to 15 days, trained on raw station data instead of smoothed reanalysis. Brightband ranks it the most accurate global model ever. Grid operators get turbine-height wind and full solar fields — aggressive passive income for the renewables crowd. Catch: data flows via Big