WeatherNext 3 Outshines Physics Models in Weather Forecasting

Google DeepMind's WeatherNext 3 has outperformed every physics model on Earth in weather forecasting, according to an independent benchmarking leaderboard run by AI weather startup Brightband. This new model updates global forecasts every hour at up to 5 km resolution for surface variables like temperature and moisture, compared to traditional models that refresh every six hours with a 25 km grid.

The WeatherNext 3 model has posted the lowest 2-meter temperature error for 26 of the last 30 days in August 2026, earning it the designation of 'regularly the most skillful out of all its peers.' Google equates an extra day of cyclone warning to a decade of meteorological progress. This is significant because a 2026 Nature paper showed WeatherNext AI models delivering three-day storm forecasts that match what previous systems managed in two.

WeatherNext 3's competitors, such as ECMWF's IFS and NOAA's GFS, are the gold-standard physics models meteorologists have relied on for decades. However, Google claims its model learns directly from real-time observations, enabling it to provide timely and more localized predictions for the weather events that impact people the most.