is a pocket-interacting molecular generative foundation model for unified 3D molecular generation via learned atomic interactions. It supports structure prediction tasks such as small-molecule docking, peptide docking, and conformation generation, as well as structure-based drug design, fragment linking/growing, PROTAC design, linear and cyclic peptide design, peptide inverse folding, and partial complex prediction or generation. The release includes Conda setup files, model-weight download instructions, Colab notebooks, sampling examples, benchmarking guides, and training/data-processing documentation.
