Introducing WeatherNext 3, our most advanced and accurate global weather AI model

Google has unveiled WeatherNext 3, an advanced AI model designed to deliver more precise, high-resolution weather forecasts. This new model leverages real-time satellite data instead of traditional physics simulations to provide hourly, localized predictions that are five times sharper than its predecessors. The enhanced forecasts are now available through Google Search, Maps, and Gemini, as well as through integration into Google Cloud projects.

The key innovation of WeatherNext 3 lies in its use of live satellite data, which is updated hourly, allowing for a much sharper and more detailed picture of weather conditions. This real-time data enables the model to generate forecasts every hour, providing information at multiple spatial resolutions ranging from 5 kilometers to 25 kilometers.

This enhanced resolution is crucial for accurately predicting rapidly changing weather events, such as storms or sudden temperature fluctuations, which can have significant impacts on agriculture, renewable energy production, and daily life.

WeatherNext 3 represents a significant leap forward in weather forecasting technology. Unlike previous models that relied on complex, supercomputer-driven physics simulations with a six-hour data lag, WeatherNext 3 directly ingests a mosaic of live, global geostationary satellite data. This allows the model to provide hourly forecasts grounded in the most recent satellite observations, offering earlier and more detailed insights into rapidly developing weather phenomena.

The model's higher resolution and faster update cycle are particularly valuable for regions that have historically lacked access to high-resolution forecasting due to the high computational costs involved. By bringing localized, high-fidelity forecasting to areas across Latin America, Africa, and Asia-Pacific, WeatherNext 3 aims to improve weather-aware decision-making for billions of people and local businesses in these regions.