NASA and IBM's open source lunar model turns 17 years of orbiter data into a foundation for lunar science
17 years of lunar data. One open-source model. 22% better ice-deposit predictions.
Why it matters
NASA and IBM released an open-source foundation model trained on Lunar Reconnaissance Orbiter imagery, demonstrating measurable improvement in polar ice detection—a concrete example of domain-specific models grounding AI in real observational data at scale.
The key facts
9 to knowLunar Foundation Model trained on ~2 million tile bundles from 17 years of LRO data
22% error reduction in polar ice deposit prediction vs. strongest baseline tested
Open-source release; no closed vendor lock-in
Domain-specific foundation model (space science, not general purpose)
No pricing, deployment details, or adoption metrics disclosed
Trained on nearly 2 million tile bundles from Lunar Reconnaissance Orbiter (17 years of data)
22 percent error reduction in polar ice deposit prediction vs. strongest baseline tested
Open-source release (license not specified in article)
Appears to be first open-source foundation model for lunar science
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: NASA and IBM have released the Lunar Foundation Model, one of the first open-source AI models for lunar science. Trained on nearly 2 million tile bundles, mostly from 17 years of Lunar Reconnaissance Orbiter data, it cuts the error in predicting polar ice deposits by up to 22 percent compared to…