Metal ions serve as essential cofactors in numerous proteins, playing a critical role in enzymatic activities and protein interactions. Given their importance, accurately identifying metal-binding sites is fundamental to understanding their biological functions, with significant implications for protein engineering and drug discovery. To address this challenge, we present SuperMetal, a generative AI framework that combines a score-based diffusion model, confidence model, and clustering mechanism to predict metal-binding sites with high accuracy and efficiency. Using zinc ions as an example, SuperMetal outperforms existing state-of-the-art tools, achieving a precision of 94% and coverage of 90%, with zinc ions localization within 0.52 ± 0.55 Å of experimentally determined positions. Furthermore, SuperMetal delivers rapid predictions and is minimally affected by increases in protein size. Notably, SuperMetal predicts metal-binding locations without needing prior knowledge of ions numbers, unlike AlphaFold3, which requires this information for its predictions. While currently trained exclusively on zinc ions, SuperMetal’s framework can be easily adapted to predict the binding sites of other metal ions by adjusting the training data.