What researchers are trying
Evo 2 uses a 40-billion-parameter long-context architecture trained on more than nine trillion nucleotides with a one-megabase context window.
Genome-scale generative biology
Can a foundation model read, predict and design biological sequence continuously from single nucleotides to megabase-scale genomes?
Why it matters
Genome-scale models could connect regulatory elements, genes, mobile elements and whole systems while supporting variant prediction and sequence design.
Evidence boundary
Generative plausibility is not equivalent to biological viability, function or safety. Long generated sequences require extensive synthesis, containment and functional review.
Organizations represented
Arc Institute · Stanford · NVIDIA
Demonstrated evidence
What has actually been shown.
- Nature publication and open model, code and integration resources reported by Arc Institute.
Unresolved questions
- What biological constraints are learned versus memorized?
- How should whole-genome designs be evaluated before synthesis?
- Can mechanistic interpretability keep pace with model scale?
Signals to watch next
- Evo Designer experiments
- Genome-scale wet-lab tests
- Interpretability tools
- Responsible-release controls
Connected evidence graph
Related BioAtlas model passports.
Evo / Evo 2
A genomic foundation model that reads and writes DNA at scale.
4/7 evidence fields documentedHyenaDNA
Long-context genomics without attention's quadratic cost.
3/7 evidence fields documentedNVIDIA BioNeMo
The GPU-accelerated toolkit that ships biology's foundation models.
2/7 evidence fields documentedProGen / ProGen2 / ProGen3
GPT-style language models that write functional proteins.
3/7 evidence fields documentedPrimary and evaluation sources