

Stanford University · v1 · 3× · vu le 08 août 2026
Evo 1 is a biological Foundation Model for DNA developed collaboratively by researchers from Arc Institute, Stanford University, and UC Berkeley. The model with 7 billion parameters is based on the StripedHyena architecture and was trained on the OpenGenome dataset comprising approximately 300 billion Tokens from prokaryotic and phage genomes. Evo 1 enables both predictive and generative design tasks at DNA, RNA, and protein levels, with a context length of up to 131,000 Tokens at single-nucleotide resolution. It was presented as a preprint in February 2024 and published in the journal Science in November 2024; the code is Open Source (Apache-2.0 license) and available on GitHub.