Protein Structure · Meta AI (FAIR)
Single-sequence folding in seconds — no MSA, no waiting.
1–30s
Run this model
About this model
ESMFold pairs the ESM-2 protein language model with a folding trunk, predicting structure directly from a single sequence. Because it needs no multiple sequence alignment it is orders of magnitude faster than MSA-based predictors and works on orphan sequences with no known homologs — the property that made ESM Atlas, a database of over 600 million predicted metagenomic structures, possible at all. Accuracy trails AlphaFold2 on targets with deep alignments and closes much of the gap on those without.
Standardized I/O contract
Every model in the Hub speaks the same contract, which is what lets the Router and the agent call any of them without special-casing.
Inputs
Protein sequence
Max residues
Outputs
All-atom coordinates in PDB format, viewable and downloadable.
pLDDT score per residue (0–100). Above 70 is generally reliable.
Global confidence in the predicted fold.
