SandboxAQ Launches AI Model for Catalyst Discovery Across the Periodic Table

SandboxAQ has launched AQCat25-EV2, a quantitative AI model designed to accelerate catalyst discovery, announced in a press release. The model is trained on the AQCat25 dataset and expands computational coverage to all industrially relevant elements by incorporating quantum spin polarization effects.

AQCat25-EV2 enables researchers to identify new catalytic materials for applications in energy, chemical production, automotive, and consumer goods. It predicts energetics with accuracy approaching that of quantum-mechanical methods while operating up to 20,000 times faster, allowing large-scale virtual screening across the entire periodic table.

Developed using NVIDIA DGX Cloud with more than 500,000 GPU hours on H100 Tensor Core GPUs, the model leverages the AQCat25 dataset of 13.5 million quantum chemistry calculations across 47,000 catalyst systems. Following its release, SandboxAQ will use the NVIDIA ALCHEMI platform to make the model broadly accessible to researchers worldwide.

AQCat25-EV2 is available on Hugging Face along with a technical manuscript detailing its architecture and validation studies.