Google DeepMind and Google Research have introduced WeatherNext 3, a new AI weather forecasting model that uses real-time satellite observations. According to Brightband’s independent live evaluations, the model is Google’s most advanced and accurate global weather model to date. Unlike models based on traditional physics simulations, WeatherNext 3 combines live geostationary satellite data with historical analyses and updates its forecasts every hour.
The model offers approximately five times higher resolution than the previous WeatherNext 2. Some surface variables, such as temperature and humidity, are forecast at 5-kilometer resolution, other surface variables at 10-kilometer resolution, and atmospheric variables such as wind speed at 25-kilometer resolution. WeatherNext 2 operated on a 25-kilometer grid at six-hour intervals. The new system also takes local topography into account and aims to produce more detailed results for rapidly changing conditions such as temperature, humidity, wind and precipitation.
WeatherNext 3 is being integrated into Google Search, Gemini, Maps, Google Maps Platform and Google Cloud. This will allow users to access forecasts through Google services, while developers and organizations will be able to use the data in their own projects. The model’s high-resolution forecasts could help expand local weather forecasting, which has previously remained limited due to high costs, particularly in Latin America, Africa and the Asia-Pacific region.
The system also offers specialized variables for renewable energy generation. It provides 100-meter wind speeds near turbine height, as well as high-resolution cloud cover and solar irradiance forecasts for solar power plants. This data aims to help grid operators and energy developers better match generation with consumer demand.
For precipitation forecasting, the model uses NASA’s satellite-based Integrated Multi-satellite Retrievals for GPM (IMERG) data, along with a global precipitation reanalysis based on satellite radar. Evaluations reported improvements of up to 60% against IMERG, 30% against MRMS and 10% against rain-gauge data in terms of the Continuous Ranked Probability Score during early forecast ranges. Google stated that local meteorological agencies should be consulted for official alerts.
Why it matters
Forecasts updated hourly and capable of distinguishing smaller areas could make it easier to monitor rapidly changing weather conditions at the local level. Integrating the model into widely used Google services is making weather data more accessible to individual users, developers and organizations, rather than limiting it to systems used only by experts. This access is particularly significant for Latin America, Africa and the Asia-Pacific region, where detailed local forecasts have remained limited due to high costs. Variables for wind, cloud cover and solar irradiance could provide grid operators and developers with more detailed data for matching renewable energy generation with consumer demand. At the same time, the performance results are based on independent live evaluations and specific metrics; local meteorological agencies continue to be the determining authorities for official weather alerts.
Background
WeatherNext is not a new name in the FikirPilot archive: we have published 2 news reports mentioning the name in the last 90 days; the latest is dated September 4, 2026.