A DNA foundation model called Evo 2 was published in Nature and can identify patterns across the genomes of more than 100,000 species and design long genomic sequences. The model accurately flags disease-causing human mutations and generated genome-length designs comparable to simple bacterial genomes. Evo 2 was trained on broad taxonomic data to capture conserved and divergent sequence features that experimental screening would take years to map. Researchers present Evo 2 as a general-purpose sequence model that compresses evolutionary information into a form usable for variant interpretation and synthetic genome design. The paper shows benchmarked performance on mutation effect prediction and design tasks; the authors discuss applications spanning functional annotation, prioritizing variants for clinical follow-up, and accelerating synthetic biology workflows.