Google upgrades WeatherNext 3 AI, boosts forecast accuracy 15%

Google announced that its new WeatherNext 3 model will power weather forecasts across Search, Maps, and Gemini, promising more accurate and personalized predictions. The update comes after years of work on AI‑driven meteorology, and the company says the model uses deep‑learning techniques to interpret massive amounts of data from satellites, weather stations, and user reports. By feeding WeatherNext 3 into the services people already use for travel, commuting, and planning, Google aims to give users a single, consistent source of weather information that is both more reliable and more tailored to local conditions. The company said the new model improves forecast accuracy by up to 15 percent over the previous generation, especially for short‑term predictions.

Weather forecasting has long relied on numerical weather prediction models that simulate atmospheric physics. Those models are computationally heavy and can struggle with complex local weather patterns. In recent years, the explosion of high‑resolution satellite imagery, global sensor networks, and cloud‑scale computing has made it possible to train machine‑learning algorithms that can learn from patterns in the data without explicit physics equations. Google’s earlier WeatherNext 1 and 2 prototypes showed promising