IBM and NASA release an open-source lunar foundation model

IBM and NASA released the NASA-IBM Lunar Foundation Model on Hugging Face together with a unified dataset that aligns over 30 layers from nine instruments across four missions, including LRO, GRAIL, and SELENE/Kaguya. The model improves ice-potential detection by 23% versus SwinV2-B and achieves up to 19% higher accuracy at 100-m resolution while using half the training data, offering a more efficient tool for crater mapping, volcanic feature analysis, and landing-site selection.