IBM and NASA have released an open-source artificial intelligence model designed specifically for lunar science, giving researchers a new tool to analyse decades of data collected from the Moon and identify geological features that could support future exploration.
The NASA-IBM Lunar Foundation Model was released on September 10 and is publicly available through Hugging Face, with its codebase also available for researchers to test and develop further. NASA describes it as among the first open-source AI foundation models built specifically for studying the Moon.
The model was developed through an ongoing collaboration between NASA, IBM Research and academic institutions. It was trained primarily on observations from NASA's Lunar Reconnaissance Orbiter, alongside a wider collection of lunar data from multiple instruments and missions.
Unlike AI systems trained for a single scientific task, the foundation model is designed to work across different types of lunar observations. Researchers can adapt it for applications such as identifying craters, mapping volcanic formations, analysing surface composition and searching for potential deposits of water ice near the lunar poles.
IBM said the model was trained using a dataset that brings together more than 30 spatially aligned data layers from nine instruments across four lunar missions. Combining information from different instruments could allow researchers to identify patterns that may be difficult to detect when datasets are analysed separately.
In testing, IBM said the model exceeded widely used approaches by up to 22% when identifying certain geographical features, including potential ice deposits, craters and volcanic formations. The performance figures were reported by IBM and NASA as part of the model's release.
Potential water ice is particularly significant for longer-term lunar exploration because it could eventually support activities ranging from producing drinking water and oxygen to generating components of rocket fuel. AI-assisted analysis could help scientists narrow down areas requiring further investigation, although model findings would still need scientific validation.
The project is part of a broader effort by NASA and IBM to apply foundation models to large scientific datasets. Their previous collaborations have included AI models for analysing Earth observation and solar data.
By making the Lunar Foundation Model open source, the organisations are also allowing researchers outside the project to experiment with the technology, fine-tune it for specialised scientific tasks and assess its performance on additional lunar datasets.
The release comes as NASA prepares for further exploration of the Moon under its Artemis programme, increasing demand for tools capable of processing large volumes of scientific and geospatial information.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.
Related Articles
