The leaderboard for atomistic modeling.

ScoreMol is an independent test lab that benchmarks machine-learning interatomic potentials, periodic and molecular quantum chemistry codes, and DFT functionals under one protocol, on held-out reference data, with the full scorecard published.

ML potentials scored: UMA-S, UMA-M, MACE-MP-0b3, Orb-v3, AIMNet2-ωB97M, CHGNet, M3GNet, EquiformerV2-OMat, SevenNet, GemNet-OC, ANI-2x, MACE-FT-Batt. Codes and packages scored: VASP, Quantum ESPRESSO, ABINIT, CASTEP, GPAW, CP2K/Quickstep, SIESTA, FHI-aims, WIEN2k, exciting, ORCA, Gaussian, Q-Chem, PySCF, Psi4, TURBOMOLE, NWChem, MOLPRO, MRCC, TeraChem, GAMESS, CFOUR, xTB, DFTB+, MOPAC.

Browse the full rankings, the coverage map, or read how scores are made.