NASA and IBM launched the NASA-IBM Lunar Foundation Model, an open-source AI tool trained on 17 years of Lunar Reconnaissance Orbiter data to analyze the Moon’s surface, detect ice, and map geological features, aiming to accelerate lunar research and future exploration.
The NASA-IBM Lunar Foundation Model represents a major leap in lunar science, combining NASA’s decades of lunar data with IBM’s AI expertise to create a tool that can rapidly analyze the Moon’s surface. The model, trained on 2 million image tiles from NASA’s Lunar Reconnaissance Orbiter (LRO), enables researchers to map craters, identify volcanic features, and estimate ice stability in permanently shadowed regions. This collaboration, announced by NASA and IBM, marks one of the first open-source AI models tailored specifically for lunar science, with the codebase available on GitHub and the model hosted on Hugging Face for public use.
The AI Model’s Development and Training
The foundation model was trained on data from NASA’s LRO, which has collected detailed observations of the Moon’s surface over 17 years. This dataset includes more than 1 million high-resolution camera images at 1-meter resolution and nearly 964,000 multispectral images at 100-meter resolution. The model also incorporated data from NASA’s GRAIL mission, Lunar Prospector, and Japan’s Selenological and Engineering Explorer. By integrating these diverse datasets, the AI can generalize across multiple scientific tasks, such as identifying craters, spotting young volcanic features, and estimating ice stability in the Moon’s permanently shadowed regions.
Kevin Murphy, chief science data officer and acting chief data and AI officer at NASA Headquarters in Washington, emphasized the model’s potential to transform lunar research. The broad knowledge they acquire through pre-training allows them to generalize across multiple scientific domains through quick fine-tuning, making foundation models both versatile and efficient in accelerating scientific research,
he said, according to NASA. The model’s ability to adapt to new tasks with minimal labeled data reduces the time and resources needed for planetary science research.
Key Capabilities and Applications
The NASA-IBM model excels in three primary tasks: detecting ice, mapping craters, and identifying volcanic features. For ice detection, the AI analyzes the Moon’s permanently shadowed regions, where temperatures remain low enough to preserve water ice for billions of years. Dark areas like the Moon’s permanently shadowed regions remain cold enough to trap and preserve ice for up to billions of years,
NASA noted. This capability is critical for planning future lunar missions, as ice could serve as a resource for sustaining human presence on the Moon.
The model also improves crater mapping, a task essential for understanding the Moon’s geological history. Traditional manual methods are time-consuming, but the AI can rapidly identify and measure craters, allowing scientists to focus on interpreting their significance. For volcanic features, the model helps locate irregular mare patches—structures that challenge existing timelines of the Moon’s thermal evolution. The AI model can also speed up crater identification and measurement, as well as analysis of unusual-looking volcanic features on the surface, to better understand the Moon’s thermal evolution,
according to sources.
Performance benchmarks show the model matched or exceeded the performance of several other strong baseline models across all evaluated tasks.
Open-Source Initiative and Future Implications
By making the model open source, NASA and IBM aim to democratize lunar research. The codebase is available on GitHub, and the model is hosted on Hugging Face, allowing scientists worldwide to experiment, refine, and adapt it for new applications. We also have to make data easier for scientists to explore and use. The NASA-IBM Lunar Foundation Model shows what’s possible when we bring AI to NASA’s petabytes of scientific data. That’s a real opportunity we see with AI: turning large-scale data into new discoveries,
Murphy said.
The initiative aligns with NASA’s broader goals for lunar exploration, including the Artemis program, which aims to establish a sustainable human presence on the Moon. By leveraging AI, researchers can process vast datasets more efficiently, accelerating discoveries and informing future missions. NASA has spent decades building an extraordinary scientific record of the Moon, but collecting data is only part of the job,
Murphy added. We also have to make data easier for scientists to explore and use.
Looking ahead, NASA and IBM plan to apply similar AI methods to future efforts aimed at studying the Moon and other parts of the solar system. The success of the lunar model demonstrates the potential of foundation models in space science, offering a blueprint for tackling complex datasets in future explorations. As the model evolves, it could become a cornerstone of lunar and planetary research, enabling scientists to uncover new insights about Earth’s nearest celestial neighbor.