AlphaFold 3

AlphaFold 3 is Google DeepMind's third-generation protein-structure model, announced in Nature in May 2024 and co-developed with Isomorphic Labs. Where AlphaFold 2 predicted single protein chains, AF3 models entire biomolecular assemblies — protein-protein, protein-DNA, protein-RNA and protein-ligand interactions — making it directly useful for drug discovery, where the protein-drug interaction is the question. The work sits on the 2024 Nobel Prize in Chemistry awarded to Demis Hassabis and John Jumper for protein structure prediction.

After months of criticism that AF3 was a restricted web-only black box, DeepMind released the code and model weights on November 11, 2024 — but with strings attached: the AlphaFold 3 Model Parameters Terms of Use permit only non-commercial use by non-commercial organizations, forbid training competing structure-prediction models on its outputs, and prohibit redistributing the parameters. Researchers at universities, nonprofits and government labs can run it freely on their own hardware; commercial drug-discovery teams cannot. Within those bounds it has become standard lab infrastructure worldwide.

• No conventional ML benchmark table; DeepMind's peer-reviewed Nature paper (2024) reports state-of-the-art accuracy on protein-ligand and protein-nucleic-acid interaction prediction (provider-run, peer-reviewed)