# Google DeepMind Launches WeatherNext 3 with Hourly 5-Kilometer Forecasts

> Google DeepMind's WeatherNext 3 is a sizable jump over last year's model: it produces hourly forecasts at 5-kilometer resolution, where the previous version worked in 25-kilometer blocks updated every six hours. It also pulls in live geostationary satellite data every hour, sidestepping the roughly six-hour lag that has long slowed traditional numerical weather prediction. The practical payoff is precipitation. DeepMind claims up to 50% more accurate rain forecasts within a day, the window that matters most for anyone deciding whether to move an event, a flight, or an emergency response. The model is about 2.4 times larger than its predecessor, and the company is wiring it into Search, Maps, Gemini, and Earth Engine. Weather is one of the clearest cases where AI models are quietly overtaking decades-old physics-based systems, not by simulating the atmosphere but by learning its patterns from data, and doing it far faster and more cheaply.

_Section: [Daily AI Updates](https://www.wortins.com/daily-ai) · Source: Google DeepMind · Published Saturday, September 5, 2026_

## Wortins' read

Google DeepMind's WeatherNext 3 is a sizable jump over last year's model: it produces hourly forecasts at 5-kilometer resolution, where the previous version worked in 25-kilometer blocks updated every six hours. It also pulls in live geostationary satellite data every hour, sidestepping the roughly six-hour lag that has long slowed traditional numerical weather prediction. The practical payoff is precipitation. DeepMind claims up to 50% more accurate rain forecasts within a day, the window that matters most for anyone deciding whether to move an event, a flight, or an emergency response. The model is about 2.4 times larger than its predecessor, and the company is wiring it into Search, Maps, Gemini, and Earth Engine. Weather is one of the clearest cases where AI models are quietly overtaking decades-old physics-based systems, not by simulating the atmosphere but by learning its patterns from data, and doing it far faster and more cheaply.

## Source

[Read the full story at Google DeepMind](https://deepmind.google/science/weathernext/)

## Related coverage

- [Ray Kurzweil Joins Subsense as Brain-Computer Interface Advisor](https://www.wortins.com/story/ray-kurzweil-joins-subsense-as-brain-computer-interface-advi-0d65bb7f) — [AI Weekly](https://aiweekly.co/ai-news-today/edition/2026-09-04)
- [Anthropic's Claude Formalizes Fermat's Last Theorem in Lean](https://www.wortins.com/story/anthropic-s-claude-formalizes-fermat-s-last-theorem-in-lean-a06abb95) — [Anthropic](https://www.anthropic.com/research/formalizing-fermats-last-theorem)
- [Elucid Raises $55M Series D for AI Cardiovascular Diagnostics Platform](https://www.wortins.com/story/elucid-raises-55m-series-d-for-ai-cardiovascular-diagnostics-03f36399) — [BioSpace](https://www.biospace.com/press-releases/elucid-raises-55-million-in-oversubscribed-series-d-financing)
- [Govee Unveils AI-Powered Smart Lighting at IFA 2026](https://www.wortins.com/story/govee-unveils-ai-powered-smart-lighting-at-ifa-2026-068839df) — [PRNewswire](https://www.prnewswire.com/news-releases/govee-ushers-in-the-next-era-of-smart-living-with-ai-powered-lighting-innovation-at-ifa-2026-302867667.html)
- [BAAI DisCo Framework Converts GitHub Repos Into Reusable AI Agent Skills](https://www.wortins.com/story/baai-disco-framework-converts-github-repos-into-reusable-ai--1c9b847a) — [AI Weekly](https://aiweekly.co/ai-news-today/edition/2026-09-04)
- [MBZUAI Releases Six Apache 2.0 K2 Horizon Open Models](https://www.wortins.com/story/mbzuai-releases-six-apache-2-0-k2-horizon-open-models-8dc864c5) — [AI Weekly](https://aiweekly.co/ai-news-today/edition/2026-09-04)

---

_Curated and written by [Wortins](https://www.wortins.com) — The daily AI briefing. Every story links to its original source; the "Wortins read" on each is our own original analysis. [About Wortins & our editorial approach](https://www.wortins.com/about)._