WeatherNext 3: Google DeepMind's Hourly High-Resolution AI Weather Model

Google DeepMind's WeatherNext 3 is a breakthrough AI model that delivers high-resolution global weather forecasts every hour. By leveraging raw satellite imagery, it provides precise local data for both everyday consumer use and complex industrial operations. The model is now integrated across Google's ecosystem and cloud platforms to enhance global weather preparedness and efficiency.

Key Points

• WeatherNext 3 is the first global AI model to provide hourly weather updates by processing raw satellite data.
• The model delivers high-resolution forecasts, reaching 5km for temperature and humidity and 10km for wind variables.
• It is designed for both consumer use via Google apps and industrial application, such as optimizing wind and solar farm operations.
• The technology improves disaster prediction accuracy, as demonstrated during Hurricane Melissa's historic landfall in Jamaica.
• Users can access the model through BigQuery, Earth Engine, and Google Cloud, or test it via the experimental Weather Lab platform.

Sentiment

The overall sentiment is mixed but constructively engaged. Hacker News generally accepts that WeatherNext 3 is a technically meaningful step and may become useful in professional or enterprise contexts, but the community is not convinced that it will automatically improve everyday local forecasts or replace established forecasting workflows. The dominant tone is cautious interest, with skepticism focused on consumer product quality, interpretability, regional performance, and access rather than on the basic idea of AI-assisted weather prediction.

In Agreement

• Hourly AI forecasts that incorporate real-time observations could be useful for hazard prediction, energy forecasting, and other operational tools.
• Using satellite and weather-station observations alongside reanalysis data is viewed as a meaningful improvement over earlier AI weather approaches.
• WeatherNext 3 may be valuable as an additional forecast source or ensemble input, especially where global model coverage is otherwise coarse.
• Specialized model outputs for phenomena such as cyclones are seen as technically interesting and potentially better than simple post-processing of gridded forecasts.
• Some users appreciate that Google is making the model visible through Weather Lab, APIs, and consumer products rather than keeping it purely internal.

Opposed

• Many commenters distrust Google weather forecasts because their everyday local forecasts have missed obvious near-term weather events.
• Several technically informed commenters argue that headline spatial resolution does not prove the model handles mesoscale weather details better than established regional NWP.
• Operational users worry that limited published fields, lack of model-member access, and weak explainability make the model hard to rely on for serious downstream workflows.
• Some commenters criticize Weather Lab's interface, regional availability, and missing meteorological views such as clearer wind direction.
• A few commenters object to the announcement's climate-risk framing as marketing, especially given concerns about the environmental cost of large AI systems.