Orb-v3: atomistic simulation at scale
About
We introduce Orb-v3, the next generation of the Orb family of universal interatomic potentials. Models in this family expand the performance-speed-memory Pareto frontier, offering near SoTA performance across a range of evaluations with a >10x reduction in latency and > 8x reduction in memory. Our experiments systematically traverse this frontier, charting the trade-off induced by roto-equivariance, conservatism and graph sparsity. Contrary to recent literature, we find that non-equivariant, non-conservative architectures can accurately model physical properties, including those which require higher-order derivatives of the potential energy surface. This model release is guided by the principle that the most valuable foundation models for atomic simulation will excel on all fronts: accuracy, latency and system size scalability. The reward for doing so is a new era of computational chemistry driven by high-throughput and mesoscale all-atom simulations.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Energy barrier prediction | Acetylene hydrogenation on Pd(111) and PdAg surfaces 1.0 (test) | -- | 32 | |
| Stability prediction | Matbench-Discovery unique structure prototypes | F1 Score0.905 | 26 | |
| Simulation Speed Estimation | Standard periodic atomic system ≈ 50 neighbors per atom, 6Å cutoff | Steps per Second77 | 23 | |
| Atomic distance prediction | Ag(111) surface adsorption | Main Chain Carbon Distance (Å)3.38 | 10 | |
| Material Discovery | Matbench-Discovery non-compliant full (test) | F1 Score88.7 | 10 | |
| Material Discovery | Matbench-Discovery 10k most stable | F1 Score96.4 | 10 | |
| Materials Stability Prediction | Matbench-Discovery | F1 Score90.5 | 8 | |
| Energy barrier prediction | PdAg surfaces | MAE0.11 | 7 | |
| Energy Prediction | Wiggle150 LAMBench | MAE (kcal/mol)11.9 | 4 | |
| Reaction Property Prediction | OC20 NEB LAMBench | MAE Activation Energy (eV)2.3 | 4 |