We prepared CryoZeta benchmark dataset to evaluate structure modeling of proteins and nucleic acids in cryo-EM maps. The dataset consists of 4 subsets organized by model variants and benchmark datasets: CryoZeta (Structures of CryoZeta on 200 test entries) CryoZeta-Interpolate (Structures of CryoZeta-Interpolate on 200 test entries) CryoZeta-Combine (Structures of CryoZeta-Combine on 200 test entries) CryoZeta_large_complex (Structures of CryoZeta on 51 large complex test entries) Each directory contains CIF format models generated by the corresponding CryoZeta variants, named as CryoZeta_[PDBID].cif, which correspond to PDB entries (e.g., CryoZeta_9mk9.cif for 9MK9 in PDB).