Harness the Power of Genomic-Scale AI: Evo 2 Now Available on Vecura

What is Evo 2?

Evo 2 is a state-of-the-art foundation model for DNA, designed for genomic-scale modeling at single-nucleotide resolution. Built on the advanced StripedHyena 2 architecture, it leverages a hybrid approach that interleaves multi-head attention with hyena convolution operators to achieve sub-quadratic scaling. This allows the model to process context windows as large as 1 million base pairs, making it highly efficient for analyzing entire viral genomes, regulatory regions, and complex gene structures.

What can users do with Evo 2 on Vecura?

With Evo 2 on Vecura, users can:

• Generate Novel Sequences: Design synthetic DNA sequences, such as novel promoters or regulatory regions, by providing a genomic prompt.
• Perform Zero-Shot Variant Effect Prediction: Assess the functional impact or pathogenicity of DNA variants (SNVs) without requiring labeled training data, achieving state-of-the-art correlation with benchmarks like BRCA1.
• Extract Genomic Embeddings: Generate high-quality, fixed-size vector representations of DNA sequences from intermediate model layers to power downstream machine learning tasks, such as exon/intron classification.
• Scale Across Contexts: Utilize various checkpoints—ranging from 1B to 40B parameters—to handle anything from standard 8,192 bp sequences up to 1-million bp genomic spans.

What the output means

The output provides sequence continuations, log-probability scores for variant assessment, or dense numerical embedding vectors. These results offer deep insights into the probabilistic structure and potential functional significance of DNA sequences.

This output should be used to support scientific decision making. It does not replace experimental validation.

Why this matters

The ability to accurately model DNA at a million-base-pair scale represents a significant leap in computational genomics. By enabling researchers to analyze entire regulatory landscapes and predict the effects of mutations with high accuracy, Evo 2 accelerates the discovery of functional genetic elements and our understanding of complex biological systems.

As genomic data generation continues to outpace manual analysis, foundation models like Evo 2 provide the necessary scalability to interpret vast amounts of sequence information. This capability is essential for modern biotechnology, drug discovery, and functional genomics, providing a powerful "in-silico" lens through which to view the genome before committing to costly experimental pipelines.

• Developed by: Arc Institute
• Source: Official GitHub Repository
• Reference: Nature (2026), "Evo 2: A genomic-scale DNA language model"