Researchers have developed a new AI model called MEHnet-MG that can predict molecular properties with high accuracy using just one inexpensive calculation. The model is trained on a dataset of multi-property labels computed at the CCSD(T) level and can derive a range of properties, including energy, optical gap, and polarizability, across nine main-group elements. In tests, the model reduced errors by a factor of 3.8 to 230 compared to other density-functional theory methods, while adding only a small amount of computational time. --- Why it matters: This matters because it could enable researchers to make more accurate predictions about molecular properties without having to perform expensive calculations. This is particularly important for large molecules where traditional methods become impractical. Source: https://arxiv.org/abs/2608.18346