MACE-MP-0
- Scope: materials, crystals, surfaces, and bulk systems — 89 elements across the periodic table.
- Training data: Materials Project DFT calculations (PBE+U functional).
- Accuracy: PBE-level. Expect roughly 0.1–0.5 eV/atom overbinding relative to experiment.
- D3 dispersion: supported and recommended for van der Waals systems (layered materials, molecular crystals).
- Energy reference: isolated-atom energies. Typical range
−1 to −15 eV/atom.
MACE-OFF
- Scope: organic molecules and drug-like compounds.
- Supported elements: H, C, N, O, F, P, S, Cl, Br, I (10 elements).
- Training data: ωB97M-D3BJ reference data (near coupled-cluster quality for organic chemistry).
- Energy reference: a different convention from MACE-MP-0. Typical range
−100 to −600 eV/atom.
Do not enable D3 with MACE-OFF
MACE-OFF is trained on ωB97M-D3BJ data — dispersion is already included. Enabling the D3 correction double-counts dispersion interactions. SimpleAtom disables the D3 toggle automatically when MACE-OFF is selected.Model sizes
Each foundation model is available in three sizes:
| Size | Speed | Accuracy | Use case |
|---|---|---|---|
small | Fastest | Least accurate | Quick screening, large systems |
medium | Balanced | Good | General use (default) |
large | Slowest | Best | Production results, publication |
Custom models
You can upload your own .model checkpoint trained with mace-torch (for example, a model fine-tuned with mace_run_train). SimpleAtom loads it with MACECalculator, runs the calculation identically to the foundation models, and can compare your model against a foundation model with agreement metrics (MAE, RMSE, R²).
Choosing a model
Is your system purely organic (only H, C, N, O, F, P, S, Cl, Br, I)?
├─ Yes → MACE-OFF
└─ No → MACE-MP-0
Do you need van der Waals / dispersion corrections?
├─ MACE-MP-0 → enable D3 dispersion
└─ MACE-OFF → do nothing (already included)Reference conventions differ
MACE-MP-0 and MACE-OFF use different energy-reference conventions, so their absolute energies differ by an order of magnitude. Never compare absolute energies across the two models — see Units & conventions.