Announcement_14

Our new preprint “Train for Accuracy, Execute at Scale: Architecture-Preserving Inference for Equivariant Atomistic Foundation Models” is now on arXiv! We present Symmetrix-XL, an inference engine that scales pretrained MACE checkpoints to 11 million atoms on a single A100 — up to 703 million atoms across 64 GPUs — with 3–5× faster inference compared to optimized baselines, no retraining required.