Forecasting just got a speed upgrade. Google launched WeatherNext 2 to deliver earlier, sharper reads on global weather. The system generates predictions up to ‘8x faster’ while offering higher-resolution views of what weather might unfold.
In the announcement, Google calls WeatherNext 2 the strongest evolution of its weather tech so far, an overhaul designed to produce more reliable forecasts and surface earlier signals of severe weather. The company said the upgraded model can map out hundreds of possible scenarios, helping users and services plan with greater confidence.
Fast forwarding forecasting
Built by teams at Google DeepMind and Google Research, the new WeatherNext 2 model is engineered to keep face with shifting skies. It delivers rapid updates and a more detailed view of what’s approaching, from sudden downpours to rising heat.
Those upgraded predictions now reach Search, Maps, Pixel Weather, Gemini, and the Google Maps Platform Weather API, meeting people wherever they look for conditions.
The added responsiveness gives users a stronger heads-up on developing weather, helping them spot trouble before it hits.
Sensing the expected and unexpected
WeatherNext 2 doesn’t settle for a single storyline. It examines hundreds of possible weather paths from the same starting conditions, giving the system a broader scope for interpreting how patterns might drift, converge, or intensify.
The model turns that expanded spread into a full 15-day outlook, giving it a steadier grip on uncertainty and the twists that come with longer-range forecasts. Testing by DeepMind and Google Research shows it provides more consistent, better-calibrated predictions than earlier WeatherNext versions.
By weighing a larger pool of outcomes, the system provides a more grounded sense of how the weather could develop, across the expected and the unexpected.
A clearer, cleaner view of how weather truly behaves
The upgraded system produces forecasts with finer detail instead of the soft, smoothed-over patterns that plagued earlier AI models. That added fidelity strengthens its handling of challenging situations, including extreme heat, powerful winds, and pressure systems that shape major storms.
It also shows a marked jump in cyclone tracking, offering path projections about a day earlier than previous versions. By better capturing how atmospheric elements interact, the model gives forecasters a more realistic read on evolving conditions.
Next-gen weather tech steps into everyday use
Google is widening access to its upgraded forecasting system by making WeatherNext 2 available through Earth Engine, BigQuery, and an early-access track on Vertex AI. These channels give scientists, developers, and public agencies direct entry to the model’s data and tools, so they can build forecasting features into their own platforms and workflows.
The company also plans to bring enhanced weather layers to Google Maps in the coming weeks, extending the system’s reach into more location-based services.
The rollout puts the model within reach of industries that use long-range weather guidance to make critical decisions, such as energy providers, insurers, emergency teams, and transit agencies. Google notes that new features and dedicated models are under active development.
As powerful storms and rapid-forming cyclones strike more often in regions unaccustomed to them, stronger forecasting systems carry real weight. AI-powered tools that can map out more realistic possibilities and support better preparation could help communities act sooner and stay safer when conditions take a dangerous turn.
Google DeepMind also detailed an upgraded version of SIMA 2 designed to interpret player intent and operate more naturally inside complex virtual worlds.