Evo 2 lands as open genome model

Evo 2 feels like the first biology model that’s trying to be infrastructure, not just a demo. The real story is less “AI writes DNA” and more “open genomic foundation model with enough scale to become a platform for downstream tools.”

• –9.3T training bases and a 1M-token context window let it capture long-range genomic relationships that older sequence models miss
• –Strong BRCA1 and variant-effect results make it immediately useful for annotation, prioritization, and research workflows
• –Full openness is a big deal for biotech teams that want to build on top of the model without vendor lock-in
• –The pathogen exclusions and safety guardrails hint at the dual-use risk that will shadow any capable genome generator
• –This is still a research model before it’s a product, but it could become the kernel for a new biology tooling stack