7-billion parameter long-context genomic transformer predicting and generating whole operons and plasmids
Evo is a biological foundation model developed by the Arc Institute, Stanford University, and Together AI. Trained at single-nucleotide resolution with a 131-kilobase context window across prokaryotes and bacteriophages, Evo enables zero-shot prediction and generative design from molecular complexes to whole genomes.
7 billion parameters trained on 300 billion nucleotides across prokaryotic and viral diversity131,072 token context window powered by the StripedHyena architecturePredicts mutation effects across protein coding and non-coding regulatory elementsGenerates novel, functional CRISPR-Cas molecular complexes and synthetic transposable elements
Evo is a biological foundation model developed by the Arc Institute, Stanford University, and Together AI. Trained at single-nucleotide resolution with a 131-kilobase context window across prokaryotes and bacteriophages, Evo enables zero-shot prediction and generative design from molecular complexes to whole genomes.
Informations confrontées à une source officielle ; pas un banc d’essai expérimental. Source · Dernière vérification : 20/09/2026, 18:50:00 UTC
Fonctionnalités clés
7 billion parameters trained on 300 billion nucleotides across prokaryotic and viral diversity
131,072 token context window powered by the StripedHyena architecture
Predicts mutation effects across protein coding and non-coding regulatory elements
Generates novel, functional CRISPR-Cas molecular complexes and synthetic transposable elements
Academic Context & Research Evidence
Biological & Workflow Fit
Biological Application
Synthetic biology, CRISPR system design, regulatory sequence engineering, and whole-genome generation
Research Workflow
Input DNA sequence (up to 131kb) -> predict fitness effects or generate de novo genomic sequences
Compute & Hardware
1x A100/H100 GPU (80GB) for inference; multi-node cluster for fine-tuning
Use this citation format when referencing Evo DNA in scientific publications and benchmark papers.
@software{evo_dna_2026,
title = {{Evo DNA}},
author = {{Arc Institute & Stanford University}},
year = {2026},
url = {https://github.com/evo-design/evo},
note = {Indexed on aibioatlas - AI for Biology and Drug Discovery}
}
Peer-Reviewed Literature & Preprints
Live scientific citations queried from Europe PMC and PubMed for Evo DNA.
⏳ Querying Europe PMC & PubMed for indexed publications...
Informations techniques et produit
Missing values mean the catalog has no recorded information. They do not mean a feature is absent.
Entry typeOpen Source Model
Access modeOpen Source
AI roleWhole-Genome Sequence Generation & Fitness Prediction
Input dataNot recorded
Output dataNot recorded
Licence conditionsApache License 2.0
Commercial eligibilityOpen weights and open-source code
Compute requirements1x A100/H100 GPU (80GB) for inference; multi-node cluster for fine-tuning
ValidationNot recorded
TypeLong-context genomic foundation model
Intended useNot recorded
CompatibilityNot recorded
ManufacturerArc Institute, Stanford University & Together AI
Biological applicationSynthetic biology, CRISPR system design, regulatory sequence engineering, and whole-genome generation
Research workflowInput DNA sequence (up to 131kb) -> predict fitness effects or generate de novo genomic sequences