NASA and IBM released the NASA-IBM Lunar Foundation Model, an open-source AI system available on Hugging Face, to help scientists study the Moon. The model excels at identifying potential ice locations and classifying craters, reducing errors by 23 percent compared to Microsoft's Swin V2-B baseline while using half the training data. It successfully identified a new crater from a SpaceX Falcon 9 rocket impact on August 5, despite overlap with an existing feature. Dr. Juan Bernabé-Moreno of IBM Research Europe noted the model's success in overcoming lunar imaging challenges, such as knife-edged shadows and feature-similar craters, by using a wedge-based training approach that separated training and testing data.

The release also includes a first-of-its-kind open-source dataset with tens of thousands of Lunar Reconnaissance Orbiter images and instrument data, enabling researchers to build their own models for future lunar exploration.