Space · Friday, September 11, 2026 · 9
sources
IBM and NASA release open-source AI model to map lunar ice and craters
IBM and NASA have released an open-source AI model trained on decades of lunar data to help identify ice deposits, craters, and volcanic features. The model, part of the Prithvi family, aims to support NASA's planned return of astronauts to the Moon in 2028. IBM Research Europe's director said it enables global researchers to explore the Moon at scale.
Bottom line — The model cut ice-location errors by 22% and crater detection by 19%, per IBM.
Go deeper (9)
- The NASA-IBM Lunar Foundation Model was trained on more than 30 layers of data from nine instruments across four missions, including NASA's Lunar Reconnaissance Orbiter and JAXA's SELENE/Kaguya, according to NASA and IBM.
- In benchmark tests, the model identified key lunar surface features up to 23% more accurately than widely used methods, NASA and IBM said.
- IBM reported that the model reduced errors in locating potential ice deposits by 22% and improved crater detection by nearly 19% while using only half the training data of comparable systems.
- The model is publicly available on Hugging Face, with its full codebase on GitHub, per NASA Science.
- NASA's Artemis program plans to land astronauts on the Moon in 2028, testing technologies for a sustained lunar presence and future Mars missions, multiple outlets noted.
- Juan Bernabe-Moreno, director of IBM Research Europe, UK and Ireland, said the model 'gives scientists a foundation to explore the Moon at scale, connecting observations across instruments.'
- Alongside the model, IBM and NASA released what they called the first unified, machine-learning-ready lunar dataset, aggregating over 30 spatially aligned layers, according to IBM.
- In a live test on August 5, the model correctly identified a new crater formed by a SpaceX Falcon 9 impact, per Cryptopolitan.
- The model is the latest addition to IBM's Prithvi family of open science models, which already covers geospatial, weather, and heliophysics applications, IBM stated.