In March 2026, Arc Institute and NVIDIA released Evo 2, billed as the largest open-source biology AI ever built. Trained on 9.3 trillion nucleotides from more than 128,000 whole genomes across all domains of life, it’s opening a new chapter in genome prediction and design.
What Evo 2 Can Do
Evo 2 can predict mutations with over 90% accuracy and design novel genomes from scratch. Researchers can use it to model how changes in DNA affect function, and to explore sequences that don’t yet exist in nature. That has huge implications for drug discovery, agriculture, and synthetic biology.
Why “Open Source” Matters
Because Evo 2 is open-source, labs and companies worldwide can run and fine-tune it without depending on a single vendor. That could accelerate both basic research and applied work in genomics, and fuel debate over safety and ethics as AI gets better at writing genetic code.
The Big Question
Scientists and commentators are already asking: how long until AI helps create synthetic life? Evo 2 doesn’t do that by itself — it’s a prediction and design tool — but it’s a major step toward understanding and eventually engineering genomes at scale. The viral coverage has put genome AI and open-source biology firmly in the mainstream spotlight.
