ESM3 & ESM C
Frontier protein language models for representation and design
What's new
ESM C ("Cambrian") open-weight models set a new bar for protein representation; a 300M ESM C rivals ESM2-650M with faster inference.
The ESM (Evolutionary Scale Modeling) family are transformer language models trained on protein sequences at the scale of life on Earth. They power everything from variant-effect prediction to generative protein design.
The family
- ESM C (Cambrian) — open-weight representation models at 300M, 600M and 6B parameters. A drop-in upgrade from ESM2 with major gains in capability and efficiency.
- ESM3 — a generative, multimodal model that reasons jointly over sequence, structure and function for controllable protein design.
Getting started
1from esm.models.esmc import ESMC2from esm.sdk.api import ESMProtein, LogitsConfig34client = ESMC.from_pretrained("esmc_300m")5protein = ESMProtein(sequence="MSEQ...")6emb = client.logits(client.encode(protein), LogitsConfig(return_embeddings=True))
Related tools
Open source
AI Models
Evo 2
A genome-scale DNA foundation model across all domains of life
Python libraryModel weightsREST API
Free for academia
AI Models
AlphaFold 3
Predict the joint structure of proteins, nucleic acids and ligands
Web appModel weightsCommand line
Open source