IBM and NASA release open-source AI model to map the Moon

IBM and NASA have released an open-source artificial intelligence model designed to help scientists analyze decades of lunar observation data, a step the agencies say could support plans to return humans to the Moon. The NASA-IBM Lunar Foundation Model became available as a free download on Hugging Face on September 10, 2026, with its full codebase published on GitHub. The model was trained on a multimodal NASA dataset spanning roughly 17 years of lunar observations. Researchers can adapt the pretrained system to a range of tasks, such as identifying ice deposits, mapping craters and studying volcanic terrain, rather than building a new, task-specific model for each question. IBM said that previously, scientists had to comb through maps and images by hand or rely on narrow, low-resolution machine-learning models that were expensive to run and often missed fine detail. According to benchmarks conducted by IBM and NASA, the model reduced errors by 23 percent when locating potential ice deposits compared with SwinV2B, a Microsoft-trained vision system widely used as a baseline for image analysis. One account of the announcement described the improvement as up to 22 percent.