MOUNTAIN VIEW, September 11 — NASA and IBM have released a new artificial intelligence model built to help scientists study the moon, part of the space agency’s broader effort to return humans to the lunar surface under the Artemis program. The NASA-IBM Lunar Foundation Model, announced Thursday, September 10, is now publicly available on Hugging Face, the open-source AI platform. It was trained on decades of lunar observation data selected by researchers at both organizations, according to a press release.
The model addresses a practical bottleneck. Until now, scientists examining the moon’s surface had to manually comb through maps and images collected by different instruments over decades, or rely on low-resolution machine learning tools built for one narrow task at a time.
The new model identifies data patterns at scale and across different instruments, saving researchers time. According to the press release, it outperforms widely used methods by up to 23% in identifying the moon’s key geographic features. Those features matter for exploration.
Craters, volcanic formations and potential ice deposits all carry crucial information for future Artemis missions, and the ability to spot them quickly and consistently could shape where crews land and what they study.
Decades of partnership, and a dataset to match
IBM and NASA have worked together for more than 60 years, dating back to their first collaboration on putting the first man on the moon. Alongside the model, scientists from both organizations built what they describe as the first open-source lunar dataset of its kind, bringing together tens of thousands of maps and images collected by nine instruments across four moon missions.
Campbell Watson, senior research manager at IBM Research, told CNET in an email that by making the model and the dataset publicly available, IBM and NASA want to provide the global scientific community with a shared foundation that researchers can adapt and apply to new questions about the moon. “This work builds on decades of publicly released lunar data collected across multiple missions, so open-sourcing the model continues that tradition of broad scientific access and collaboration,” Watson said.
The lunar model is not the only open-source AI tool from NASA and IBM on Hugging Face. The platform also hosts the Prithvi family of models for weather, climate and earth observation. Researchers can build on the NASA-IBM model to make new discoveries rather than creating an AI model from scratch each time.
The open-source approach is intended to let the global scientific community adapt and apply the model to new lunar research questions. The announcement was covered by CNET, with the story published on September 10, 2026, at 8:05 a.m. ET.
What to watch next is how quickly outside researchers put the model to work, and whether the patterns it surfaces help shape the science that will guide Artemis crews back to the moon.
