NASA-IBM Lunar Foundation Model Goes Open Source with a 2M-Tile Dataset and 22% Lower Ice-Mapping Error
StorageReview, Monday, September 14th, 2026
StorageReview reports IBM and NASA open-sourced a lunar foundation model trained on two million tiles from nine instruments.
IBM and NASA have released the NASA-IBM Lunar Foundation Model as open source, one of the first publicly available foundation models built specifically for scientific study of the Moon.
The release includes model weights, a technical report and the training dataset on Hugging Face, covering roughly two million tiles with 30 data layers drawn from nine lunar instruments.
On downstream ice-mapping tasks the model reduces error by about 22% relative to prior approaches. StorageReview highlights the open data release as the more durable contribution, since the multi-instrument tile corpus is reusable well beyond the published model.