Google DeepMind and Isomorphic Labs' model predicting structures of proteins, nucleic acids, ligands and their complexes.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
AlphaFold 3 is a biomolecular structure prediction model from Google DeepMind and Isomorphic Labs. It predicts the 3D structure and interactions of proteins, DNA, RNA, ligands and other molecules, extending AlphaFold 2, which won CASP14 and predicted single protein structures. The method is described in a 2024 Nature paper, "Accurate structure prediction of biomolecular interactions with AlphaFold 3".
It is available in several ways. AlphaFold Server offers free, non-commercial web access with a more limited set of ligands and covalent modifications. The GitHub repository provides the inference pipeline code (Apache-2.0), while model weights are provided directly by Google for academic use under separate terms of use. For commercial use, AlphaFold 3 is offered through Google Cloud. Running it locally requires a GPU and large genetic databases; the data pipeline step is CPU-only and slow.
Limits: the weights may only be used if obtained directly from Google and under their terms, which restrict commercial use. A YouTube comparison reported it lost to IBM's MAMMAL on 5 of 7 antibody-target binding tests, which the creator attributed to difficulty with intrinsically disordered, flexible regions. The related AlphaFold Protein Structure Database, built with EMBL-EBI, holds over 200 million AlphaFold-predicted structures and is free to use.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Protein structure prediction model used as the comparison baseline. It lost to MAMMAL on 5 of 7 antibody-target binding tests, which the creator attributes to AlphaFold struggling with intrinsically disordered, flexible regions.
Closest neighbours by meaning, not just by category.
Model · Google DeepMind
Google DeepMind diffusion pipeline for de novo enzyme design by motif scaffolding, with open code and downloadable weights.
Research · Google DeepMind
Searchable database of AlphaGenome predicted effects for every possible single-nucleotide change in the human genome.
App · Google
Google suite of science tools whose "science skills" connect Gemini to 30+ life science databases such as UniProt and AlphaFold.
Model · Hugging Face
Open 14.8B language model from Hugging Face that reads raw DNA in chat and answers in English, built on Qwen3-14B.
Model
Multimodal biomedical foundation model for drug discovery across small molecules, proteins, antibodies and gene expression.
Agent · Google DeepMind
Gemini-based multi-agent system from Google that generates, debates and ranks scientific hypotheses for researchers.
