NASA and IBM's open source lunar model turns 17 years of orbiter data into a foundation for lunar science

NASA and IBM Research have released the NASA-IBM Lunar Foundation Model, an open-source model trained on nearly 2 million tile bundles from 17 years of Lunar Reconnaissance Orbiter observations. The pretrained model matched or beat baselines on crater detection, ice deposit prediction, and IMP segmentation, cutting ice prediction error by up to 22 percent versus SwinV2-B. It is not suited for geodetic positioning, and researchers position it as a foundation for downstream tasks rather than a replacement for physical measurements.