Model family

Evo 2

Family overview — summarizes releases; licenses and availability belong to each release.

Evo 2 is a family of DNA language models from Arc Institute, developed with NVIDIA and collaborators at institutions including Stanford University, UCSF, UC Berkeley, Goodfire, and the University of Washington. The models use the StripedHyena 2 architecture, work at single-nucleotide resolution with context of up to 1 million base pairs, and were trained on the OpenGenome2 dataset. Checkpoints range from 1B to 40B parameters; a 20B checkpoint derived from the 40B model was added in February 2026.1 3 2

Website: Evo 2 page (Arc Institute) (external site: arcinstitute.org)
Repository: GitHub repository (external site: github.com)
Model hub: Evo 2 checkpoints (Hugging Face) (external site: huggingface.co)
Dataset hub: OpenGenome2 training dataset (external site: huggingface.co)
Paper: Nature paper (doi:10.1038/s41586-026-10176-5) (external site: doi.org)

Last reviewedSep 29, 2026Entry updated Sep 29, 2026Documented release Feb 2025

Releases assessed

Each release is assessed on its own. This family has no family-wide openness label.

Releases in the Evo 2 family
Release	Weights	Tier (USASI rubric v0.1)	License	Released
Evo 2 20B	Weights: Public	Model-disclosure tier (USASI rubric v0.1): Open-weight	Apache-2.0, Apache-2.0	Feb 28, 2026
Evo 2 40B	Weights: Public	Model-disclosure tier (USASI rubric v0.1): Open-stack	Apache-2.0, Apache-2.0, Apache-2.0	Feb 2025

What it is useful for

The repository documents scoring the likelihood of DNA sequences (for example zero-shot BRCA1 variant effect prediction), extracting embeddings for downstream classifiers, and generating DNA sequences from prompts.1

Run and use notes

Documented facts only. No hardware or performance claims are made without a cited source and stated assumptions.

Evo 2 is also offered through an NVIDIA-hosted API and as an NVIDIA NIM container, and NVIDIA BioNeMo provides an Evo 2 fine-tuning tutorial, per the repository README.1

Organization context

Provenance and derivatives

Pretrained by Arc Institute and collaborators on OpenGenome2, a dataset of about 8.8 trillion base pairs assembled from public genome and sequence databases, using the Savanna training framework.1 4

Derived from: OpenGenome2 (external site: huggingface.co) — Pretraining dataset published by Arc Institute.

U.S. eligibility

Project eligibility rests on documented governing or maintaining entities, not on contributors.

Eligible · basis: U.S.-governed project

Evo 2 is published and maintained by Arc Institute in its GitHub and Hugging Face organizations. Arc Institute describes itself as an independent nonprofit research organization in Palo Alto, California, and the IRS lists Arc Research Institute at its Palo Alto address as a 501(c)(3) organization. Eligibility of each release is assessed on the release record.1 5 6

Assessed Sep 29, 2026

Sources

1.
ArcInstitute/evo2 README (external site: github.com)
Arc Institute · Repository · accessed Sep 29, 2026
2.
Release v0.5.0 · ArcInstitute/evo2 (external site: github.com)
Arc Institute · Release notes · published Feb 28, 2026 · accessed Sep 29, 2026
3.
Evo 2: One Year Later (external site: arcinstitute.org)
Arc Institute · Announcement · published Mar 4, 2026 · accessed Sep 29, 2026
4.
arcinstitute/opengenome2 dataset card (external site: huggingface.co)
Arc Institute · Dataset card · accessed Sep 29, 2026
5.
About | Arc Institute (external site: arcinstitute.org)
Arc Institute · Official page · accessed Sep 29, 2026
6.
Exempt Organizations Business Master File Extract - California (eo_ca.csv) (external site: irs.gov)
Internal Revenue Service · Filing · accessed Sep 29, 2026

Report a correction Edit this entry

This listing is not an endorsement, a safety assessment, or a federal approval.