NASA and IBM Research have released the NASA-IBM Lunar Foundation Model, an open-source multimodal AI model designed to analyze diverse lunar datasets. Developed using the SomBench dataset, the model was pretrained from scratch on nearly two million co-registered data bundles. These bundles encompass 11 distinct modalities—including imagery, topography, illumination geometry, mineralogy, radar, and gravity—spanning spatial scales from meter-scale Narrow Angle Camera (NAC) observations to regional-scale Wide Angle Camera (WAC) data (1m to 20km per pixel). The Universities Space Research Association (USRA) contributed critical planetary science expertise, specifically assisting in connecting physical lunar characteristics to model development and helping create the high-resolution LROC NAC crater benchmark, which includes over 49,000 manually identified craters. Performance evaluations were conducted across three downstream benchmarks: crater detection at regional and meter scales, segmentation of irregular mare patches (IMPs), and regression of lunar polar ice prospectivity. Results indicate the model matches or exceeds the performance of models using ImageNet pretraining or those initialized without lunar-specific pretraining. Notably, the model demonstrated high label efficiency in crater detection tasks, proving that its learned representations can significantly reduce the volume of task-specific labeled data required for lunar machine learning applications.