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

NASA and IBM Research, working with several academic institutions, have released the NASA-IBM Lunar Foundation Model, which the organizations describe as one of the first open-source foundation models for lunar science. The model is publicly available on Hugging Face, its code is on GitHub, and it is integrated into the open-source toolkit TerraTorch. The team also released ML-ready pretraining datasets and benchmark collections.

The model was trained from scratch using SomBench, described as the largest co-registered multimodal lunar corpus to date. It contains nearly 2 million tile bundles across 11 modalities and two spatial scales. About 1 million high-resolution images come from the Narrow Angle Camera at roughly 1 meter per pixel, and just under 964,000 multispectral images come from the Wide Angle Camera at 100 meters per pixel. The bulk of the data comes from 17 years of Lunar Reconnaissance Orbiter observations; NASA says that mission's data volume exceeds that of all other NASA planetary missions combined. GRAIL gravity data, Lunar Prospector hydrogen data, and JAXA's Kaguya/SELENE mineralogy data rounded out the collection, totaling more than 30 spatially aligned data layers from nine instruments and four missions.

The model is based on TerraMind, a multimodal Earth observation model, but researchers trained it from scratch rather than fine-tuning. Each tile includes imaging geometry as explicit context, such as illumination angles, sun position and tile extent.

On tests covering crater detection at 100-meter and 1-meter scales, polar ice deposit prediction, and segmentation of Irregular Mare Patches, the pretrained model matched or beat common baselines and an architecturally identical control model with random initialization, according to the technical report. IBM says the model cut ice prediction error by up to 22 percent compared with the best baseline, SwinV2-B.

According to the researchers, the model is not suited for absolute geodetic positioning; in generation tests latitude and longitude were off by dozens of degrees in some cases. They see it as a reusable foundation for downstream tasks, not a replacement for physical measurement instruments. Controlled experiments isolating each innovation are still pending, and some test datasets are small.

The model is part of the NASA-IBM AI for Science collaboration under a Space Act Agreement since early 2022. In August 2023 the organizations released the first Prithvi model on Hugging Face, trained on Landsat and Sentinel-2 imagery. IBM developed TerraMind in 2025 with ESA and Forschungszentrum Jülich for Earth observation.