Nvidia and Arc Institute Launch Evo 2, the Largest AI Model for Biological Research

Evo 2, trained on 9.3 trillion DNA base pairs from 128,000 species, offers groundbreaking tools for genome analysis and design across medicine, agriculture, and environmental science.

Evo 2 is a 40-billion-parameter AI model designed to analyze and generate DNA sequences, enabling advancements in synthetic biology and genome design.
Trained on 9.3 trillion DNA base pairs from over 128,000 species, it spans all domains of life, including humans, plants, and bacteria.
The model demonstrated 90% accuracy in identifying harmful mutations in the BRCA1 gene, linked to breast cancer, and excels in analyzing non-coding mutations.
Built using Nvidia’s DGX Cloud and StripedHyena 2 architecture, Evo 2 processes DNA sequences up to 1 million base pairs, uncovering distant genetic relationships.
Freely available on Nvidia's BioNeMo platform, Evo 2 is open-source, providing researchers worldwide with tools for applications in healthcare, agriculture, and environmental challenges.