Evo 2 7B
Arc Institute's open 7B genomic foundation model that reads and designs DNA.
About
Evo 2 is a genomic foundation model from Arc Institute with NVIDIA, Stanford, UC Berkeley, and UCSF, announced February 19, 2025. It treats DNA as a language: the 7B version trained on 2.4 trillion tokens (the 40B flagship on 9.3 trillion nucleotides across 128,000+ genomes), using the StripedHyena 2 architecture with sequence contexts up to a million base pairs.
Weights, training code, inference code, and the OpenGenome2 dataset are all openly released, and the model both predicts variant effects (over 90% accuracy on BRCA1 mutation impact, per Arc) and designs new genomic sequences. It is a research tool for genomics and synthetic biology, not a general chatbot, and biological-design capabilities carry obvious dual-use considerations.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
- 90%+ accuracy predicting BRCA1 mutation impact (Arc-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Landmark open genomics model; published in Nature with experimental validation; Fully open: weights, code, training data (OpenGenome2); Specialist research tool; biological design capabilities carry dual-use considerations. License: Open weights, code, and data (see project for terms).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Landmark open genomics model; published in Nature with experimental validation
- · Fully open: weights, code, training data (OpenGenome2)
- · Specialist research tool; biological design capabilities carry dual-use considerations