Google has released WeatherNext 3, an AI weather model that outperforms traditional forecasting systems. According to TechCrunch, the model beats the U.S. National Weather Service and European forecasting systems in accuracy tests.
WeatherNext 3 ranks highest on Operational WeatherBench, a tool that compares AI forecasts across temperature, windspeed, and humidity metrics. It also outperforms deep-learning models from Microsoft, Nvidia, and others.
The model's advantage extends beyond accuracy. It predicts down to 5 kilometers, compared to the previous 15-to-25 km range, and delivers hourly forecasts instead of six-hour intervals. Rain predictions improved by 60% over the previous version. This means users can receive specific timing for weather events rather than broad windows.
Google plans to integrate WeatherNext 3 into Search, Google Maps, and Gemini, making the forecasts available to billions of users worldwide.
Weather forecasting has long relied on expensive government supercomputers. AI has changed that calculation. A startup or smaller nation can now run comparable forecasts on cloud platforms. For India, where monsoon predictions affect agricultural yields and urban planning, this technology could shift disaster preparedness and crop management. Google's decision to embed the model in consumer products means hyperlocal forecasts could reach anyone with a smartphone.
A Google staff engineer told TechCrunch this is the first time core weather variables will "feed and power a lot of the Google products", indicating that AI weather has become infrastructure rather than a feature.
Forgetting an umbrella will soon lose its excuse. The AI will know.
