NASA and IBM have jointly released an open-source artificial-intelligence model, dubbed “Surya,” that aims to improve the forecasting of solar flares and other space-weather events that threaten satellites, power grids and communications systems. The model—now available on the Hugging Face repository—was trained on nine years of high-resolution imagery from NASA’s Solar Dynamics Observatory, which has monitored the Sun since 2010. Testing shows Surya can predict a solar flare up to two hours in advance and improves classification accuracy by 16% compared with existing tools, according to IBM researchers. The additional lead time could give grid operators and satellite owners critical minutes to protect equipment from geomagnetic storms that, in a worst-case scenario, could inflict an estimated US$2.4 trillion in global economic losses over five years, Lloyd’s has warned. IBM said it is exploring longer-range forecasts while scientists evaluate the model against data from the current Solar Cycle 25. By releasing the code and training weights publicly, the partners hope researchers and government agencies will refine the system and accelerate the development of early-warning capabilities for space weather.