FrontierSeptember 6, 2026via The Decoder
Google's WeatherNext 3 ditches physics simulations and learns weather directly from live satellite data
Why it matters
A major capability milestone: Google's weather model abandons traditional physics simulations for pure ML learning from satellite data, achieving 5km resolution forecasts and extending forecast accuracy to underserved regions. This represents a shift in how frontier labs approach domain-specific prediction problems—pure data-to-prediction learning rather than hybrid physics-informed approaches.
Key signals
- WeatherNext 3 skips physics simulations entirely, learns directly from real-time satellite data
- Produces hourly forecasts at 5-kilometer resolution
- 5x more detailed than WeatherNext 2 predecessor
- Targets forecast gaps in Africa, Latin America, Asia-Pacific
- Released by Google Research and DeepMind collaboration
- Model capability breakthrough in weather domain
The hook
Google ditches 300 years of physics. WeatherNext 3 learns weather directly from satellite data—5x finer resolution, global coverage gap closed.
Google Research and DeepMind are releasing WeatherNext 3, a weather model that skips traditional physics simulations and learns directly from real-time satellite data. It produces hourly forecasts at up to five-kilometer resolution, five times more detailed than its predecessor. Google says regions …