# Evo 2: an AI that reads and writes genomes across all life
> Arc Institute's 40B-parameter DNA foundation model predicts disease mutations and designs bacterial-scale genomes, the design-tool side of the biosecurity debate

**Meta:** type: event · date: 2026-03-19 · heads: The Long Game, What They're Not Saying · 8 takes · 3 lenses · 2 regions

## Summary

[Arc Institute](/ko/entity/arc-institute) released [Evo 2](/ko/entity/evo2), a DNA foundation model trained on over 9 trillion nucleotides from more than 100,000 species across the tree of life, published in Nature in March 2026. At 40 billion parameters with a 1-megabase context and single-nucleotide resolution, Evo 2 both predicts disease-causing [mutations](/ko/entity/generative-biology) and generatively designs genomes as long as simple bacteria. It was built with [NVIDIA](/ko/entity/nvidia) (hosted on BioNeMo) and researchers at Stanford, UC Berkeley and UCSF. Evo 2 is the design-tool half of the [biosecurity-screening debate](/ko/n/dna-synthesis-screening-letter-2026): a model that can write novel functional DNA is exactly what worries the labs pushing mandatory [synthesis screening](/ko/entity/dna-synthesis), since AI-generated sequences may evade conventional motif checks.

## By the numbers

- 40B, model parameters.
- >9 trillion, nucleotides in training data.
- >100,000, species spanned (all domains of life).
- 1 megabase, context length, single-nucleotide resolution.
- Mar 2026, published in Nature; weights and code released.

## Why it matters

Evo 2 marks generative biology crossing into genome-scale design, the predictive payoff (variant effect, drug targets) and the dual-use risk (designing novel sequences) arrive in the same open model. It is the concrete reason the AI×biology screening fight is live now, not hypothetical.

## What to watch

- Independent validation of Evo 2's variant-effect and design claims.
- Whether release norms tighten for generative-biology models.
- Integration of design tools like Evo 2 into drug-discovery pipelines.

## Regional takes (batched by bias / lens)

### unlabelled
- **Arc Institute (Evo 2 release)** (United States, en) — Arc Institute's own announcement of Evo 2, a DNA foundation model trained on >9 trillion nucleotides from over 100,000 species across the tree of life; 40B parameters, 1-megabase context, single-nucleotide resolution, able to identify disease-causing mutations and generate genomes the length of simple bacteria.
  Source: https://arcinstitute.org/news/evo2
- **Nature (Evo 2 paper)** (United Kingdom, en) — The Evo 2 model and tooling page (published in Nature, Mar 2026): the largest AI model for biology to date, built with NVIDIA and researchers at Stanford, UC Berkeley and UCSF; both predictive (variant effect) and generative (sequence design) capabilities, weights and code released.
  Source: https://arcinstitute.org/tools/evo
- **Phys.org** (United States, en) — 
  Source: https://phys.org/news/2026-03-evo-ai-genetic-code-domains.html
- **Freethink** (United States, en) — 
  Source: https://www.freethink.com/biotech/evo-2-generative-biology
- **Goodfire (interpretability)** (United States, en) — 
  Source: https://www.goodfire.ai/research/interpreting-evo-2
- **Arc Institute (one year later)** (United States, en) — 
  Source: https://arcinstitute.org/news/evo-2-one-year-later

### scientific / sector
- **GEN (Genetic Engineering News)** (United States, en) — Reports Evo 2 as designing genetic code across all domains of life and stresses the dual-use tension: a generative model that writes novel DNA is exactly the 'biological design tool' the biosecurity-screening debate centres on, raising containment and release-norm questions.
  > "Arc Institute's AI Model Evo 2 Designs the Genetic Code Across All Domains of Life."
  Source: https://www.genengnews.com/topics/genome-editing/arc-institutes-ai-model-designs-the-genetic-code-across-all-domains-of-life/

### infrastructure / vendor
- **NVIDIA (BioNeMo)** (United States, en) — NVIDIA's account of hosting Evo 2 on its BioNeMo platform, framing genome-scale foundation models as a new compute workload and positioning its hardware at the centre of the AI-for-biology stack.
  > "AI for Biomolecular Sciences Now Available via NVIDIA BioNeMo."
  Source: https://blogs.nvidia.com/blog/evo-2-biomolecular-ai/

## Across the graph
- Related: [[dna-synthesis-screening-letter-2026]], [[isomorphic-labs-series-b-2026]]
- Entities: Evo2, Arc Institute, Generative Biology, AI Drug Design, Nvidia, Dna Synthesis

---
Canonical: https://rbtfl.xyz/ko/n/evo2-generative-genome-2026