PeptideCLM-23M-all — interactive explorer

aaronfeller/PeptideCLM-23M-all is a 23M-parameter RoFormer masked language model pretrained on 10.8M peptides and 12.6M small molecules, reading molecules as SMILES strings rather than amino-acid letters — so it also covers non-canonical residues, cyclisation and modified backbones.

It is an encoder: it fills in masked chemistry and produces embeddings. It does not chat and it does not generate text.

Hide one chemical token and see what the model puts back. Tokenize first to pick a position.

SMILES

Examples

SPE tokens

Position to mask

0 64

Predictions to show

1 20

Most likely tokens

Structure

Model: MIT-licensed, by Aaron Feller et al. Tokenizer vendored from AaronFeller/PeptideCLM. Research and educational use — not a clinical or diagnostic tool.