EvolutionaryScale

weights

open(ungated esm3_sm_open_v1.pth with structure and function decoder weights alongside)

data

documented-not-released(composition described by scale -- 2.78B natural proteins, 3.15B sequences with synthetic augmentation, 236M structures, 539M function-annotated proteins, 771B tokens -- but no corpus or reconstruction scripts released)

code

partial(inference, generation and tutorial code in the esm package

license

MIT(OSI

ESM3's open 1.4B checkpoint can be downloaded and run locally without an access request, under the MIT license stated in the repository's LICENSE.md and on the card. The training corpus is described in detail but not released, and the published code covers inference, generation and tutorials rather than pretraining. The 7B and 98B models are reachable only through the hosted API.

• https://huggingface.co/api/models/EvolutionaryScale/esm3-sm-open-v1 recorded 2026-10-21

  Redirects to huggingface.co/api/models/biohub/esm3-sm-open-v1; gated: false; private: false; no license tag and no cardData license key; data/weights/esm3_sm_open_v1.pth, esm3_structure_decoder_v0.pth and esm3_function_decoder_v0.pth present.

• https://huggingface.co/biohub/esm3-sm-open-v1/raw/main/README.md recorded 2026-10-21

  Front matter declares only language and tags, no license. The body says "This repository is under a MIT license" and links the esm repository LICENSE.md; training is described as 2.78 billion natural proteins, with no corpus released.

• https://raw.githubusercontent.com/Biohub/esm/main/_assets/ESM3_README.md recorded 2026-10-21

  "These models are available under the MIT license." Code covers local inference and generation and a tutorials cookbook; the 7B and 98B tiers are reachable only through the hosted platform.

• https://raw.githubusercontent.com/Biohub/esm/main/LICENSE.md recorded 2026-10-21

  "License (MIT)", Copyright 2026 Chan Zuckerberg Biohub, Inc., with the standard MIT permission grant.

Adoption is measured as trailing-30-day Hugging Face downloads of biohub/esm3-sm-open-v1, the one openly distributed checkpoint. The larger 7B and 98B tiers are served only through the hosted Forge API, which publishes no artifact count, so they sit outside the counted scope rather than being folded into it.

• https://huggingface.co/api/models/biohub/esm3-sm-open-v1 recorded 2026-09-10

  downloads: 2772; likes: 330.

Two attributes are read directly off the publisher's description: three joint sequence, structure and function tracks, and generation from a partial prompt. Adaptation is not credited, because the release documents how to prompt and sample rather than how to fine-tune, and the repository's fine-tuning guidance targets ESMC rather than ESM-3.

• https://huggingface.co/biohub/esm3-sm-open-v1/raw/main/README.md recorded 2026-09-10

  "a generative model capable of designing proteins conditioned on partial prompts of sequence, structure and function"; usage points to the esm repository notebooks, with no fine-tuning entry point.

• https://raw.githubusercontent.com/Biohub/esm/main/_assets/ESM3_README.md recorded 2026-09-10

  "able to jointly reason across three fundamental biological properties of proteins: sequence, structure, and function ... represented as tracks of discrete tokens at the input and output"; "ESM3 is a generative masked language model. You can prompt it with partial sequence, structure, and function keywords, and iteratively sample masked positions".