GNoME

Google DeepMind · 29 November 2023

Graph networks predicted 2.2 million new crystals, 380,000 of them stable; 736 later independently synthesized.

Active-learning loop between a graph neural network and DFT calculations lifted stability-prediction precision from about 50% to 80%. Released 380,000 stable structures to the Materials Project. Berkeley Lab's A-Lab reported autonomous synthesis in a companion paper.

Date: Wednesday, 29 November 2023
Lab: Google DeepMind
Kind: model
Access: research preview

Figures

Measure	Value	Measured by
New crystals predicted	2.2M 380,000 stable; prior computational total 28,000	company

Published in Nature 2023-11-29. Counts are model predictions; DeepMind reports 736 independently synthesized by outside groups.

Sources

deepmind.google/blog/millions-of-new-materials-discovered-with-deep-learning/

This record was checked against its sources on 6 October 2026. How we check