Harness the Power of Genomic Generation with Evo 2 40B on Vecura

What is Evo 2 40B?

Evo 2 40B is a powerful genomic foundation model developed by the Arc Institute that leverages a 40-billion parameter StripedHyena 2 architecture. Trained on nearly 9 trillion nucleotides across all domains of life—including bacteria, archaea, eukaryotes, and viruses—the model specializes in generating biologically coherent DNA sequences. By predicting continuations from seed nucleotide strings, it bridges the gap between raw computational sequence modeling and synthetic biology research.

It helps users generate synthetic DNA sequences that reflect learned evolutionary patterns without the need for manual, base-by-base construction. It is especially useful for researchers looking to design promoters, coding region starts, or intergenic elements for downstream wet-lab validation and synthetic biology experiments.

What can users do with Evo 2 40B on Vecura?

With Evo 2 40B on Vecura, users can:

• Generate biologically plausible DNA continuations from specific seed sequences to explore genomic sequence space.
• Augment existing training datasets to improve the performance of smaller, task-specific genomic models.
• Conduct local sequence neighborhood exploration around known functional elements to understand evolutionary variations.
• Obtain advanced metrics like raw logit values or sampled probabilities to perform likelihood-based scoring of generated sequences.

What the output means

The output provides a generated DNA nucleotide string that extends the user's input sequence. When specific flags are enabled, users can also receive raw logit values for every position or the probability confidence assigned to each chosen nucleotide.

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

Why this matters

The ability to generate high-fidelity synthetic DNA sequences at a scale of up to one million base pairs represents a significant leap in computational genomics. By encoding the "grammar" of life found across the entire evolutionary spectrum, models like Evo 2 40B allow researchers to simulate complex genomic regions that would otherwise be difficult to design from scratch, accelerating the design-build-test cycle in synthetic biology.

Integrating this technology into accessible platforms like Vecura removes the significant barrier of managing the specialized, high-performance hardware (NVIDIA Hopper architecture) usually required to run such large-scale models, democratizing access to cutting-edge genomic design tools.

• Developed by: Arc Institute
• Source: Arc Institute - Evo 2 GitHub
• Reference: Genome modeling and design across all domains of life with Evo 2 (Science 2025)