NASA and IBM release open-source lunar foundation model trained on 17 years of orbiter data

October 4, 2026 at 03:10 PM2 min

This digest was compiled by AI from multiple sources — links to the originals are below.

NASA and IBM Research released the NASA-IBM Lunar Foundation Model, an open-source foundation model for lunar science. The model was trained on nearly 2 million tile bundles from 17 years of Lunar Reconnaissance Orbiter observations. It is especially strong at predicting ice deposits at the poles and detecting craters.

Key Facts

The model was trained from scratch on SomBench, the largest co-registered multimodal lunar corpus to date, containing 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, while just under 964,000 multispectral images come from the Wide Angle Camera at 100 meters per pixel.
The bulk of the training data comes from 17 years of observations by the Lunar Reconnaissance Orbiter, whose data volume exceeds that of all other NASA planetary missions combined.
The dataset brings together more than 30 spatially aligned data layers from nine instruments and four missions, including GRAIL, Lunar Prospector, and JAXA's Kaguya/SELENE probe.
The model receives imaging geometry as explicit context, including illumination angles, sun position, and tile extent, rather than inferring lighting conditions from raw pixels.

Model Training and Data

The team trained the model from scratch using SomBench, which they describe as the largest co-registered multimodal lunar corpus to date. SomBench 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, while 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 observations by the Lunar Reconnaissance Orbiter, whose data volume exceeds that of all other NASA planetary missions combined. Data from the GRAIL mission, Lunar Prospector, and JAXA's Kaguya/SELENE probe rounded out the collection, bringing together more than 30 spatially aligned data layers from nine instruments and four missions.

Architecture and Capabilities

The model is based on TerraMind, a multimodal Earth observation model, but the researchers trained it from scratch instead of fine-tuning it. For each tile, the model receives the imaging geometry as explicit context, including illumination angles, sun position, and tile extent. On the Moon, lighting geometry shapes how the surface looks far more than the surface's actual properties do, so the team feeds that information directly to the model. The model also learns from high-resolution and coarse imagery together in a single training run, capturing both fine details and large-scale context. It is especially strong at predicting ice deposits at the poles and detecting craters.