Evo 2
A genome-scale DNA foundation model across all domains of life
What's new
Published in Nature in 2026: a 40B-parameter model with a 1M-token context trained on 9 trillion DNA bases; fully open including the OpenGenome2 dataset.
Evo 2 is a DNA foundation model trained on 9 trillion nucleotides spanning all domains of life, with a 1 million-token context window at single-nucleotide resolution.
Highlights
- Predicts the functional impact of genetic variation — including noncoding and clinically significant BRCA1 variants — without task-specific fine-tuning.
- Learns interpretable features such as exon–intron boundaries and transcription-factor binding sites.
- Generates coherent genome-scale sequences and is fully open: parameters, training/inference code and the OpenGenome2 dataset.
Related tools
Open source
AI Models
ESM3 & ESM C
Frontier protein language models for representation and design
Python libraryREST APIModel weights
Free for academia
Single-Cell & Spatial
Arc State
A virtual cell model predicting perturbation responses
Python libraryModel weights
Free