Google launches WeatherNext 3: AI weather model with real-time satellite data and hourly updates
ymatias · x · 2026-09-04
Google DeepMind and Google Research officially launched WeatherNext 3, their most accurate AI weather model yet. It now learns directly from real-time geostationary satellite observations, bypassing the 6-hour data lag of traditional numerical weather prediction, generating hourly forecasts at up to 5-km resolution. Trained on NASA's IMERG and Google's own satellite-radar-based precipitation reanalysis, it achieves up to 60% CRPS improvement against IMERG, and adds precise precipitation forecasting and clean-energy variables.
More from Research
- Looped transformer debate revisited: Sparse Universal Transformer (EMNLP 2023) got there first — Yikang_Shen · 2026-09-04
- Matrices are graphs and graphs are matrices: linear algebra's most undervalued fact — TivadarDanka · 2026-09-04
- Simulation physics gaps teach robots tricks that fail in the real world — binarybits · 2026-09-04
- How robot startups scrape for data: free cleanings, exoskeletons, sim limits — binarybits · 2026-09-04
- Google Research uses transfer learning to improve genomic prediction for underrepresented populations — Google Research · 2026-09-04
- AI benchmarks may understate AI by 82%: routing across 44 LLMs cuts errors 46% — CodeByPoonam · 2026-09-04