September 19, 2026
Anthropic puts AI into real labs: a competition will test more than 5,000 protein designs
On September 17, Anthropic launched the Life Sciences Verification Program in beta for teams and institutions. After verifying researchers' qualifications, safety measures, and ethical oversight, an organization receives access to Mythos 5.1, Opus 5, and Sonnet 5 through Claude Science, Claude.ai, Claude Code, and the API. Anthropic also open-sourced optimizations for more than 30 biomolecular models, making them 4 times faster on average.
Anthropic is already running physical experiments in its own Bay Area biology lab and with external partners. Life sciences lead Eric Kauderer-Abrams calls fundamental biology, rather than drug discovery, the main focus.
A shift in scale. In April, Anthropic acquired Coefficient Bio, and on September 16, Novo Nordisk announced it was testing Claude Science in drug research and development. Anthropic has now confirmed its own lab work and opened a formal path for organizations to access the models.
Terms. Applications for LSVP are submitted through a form, but submitting it does not guarantee access. The program works through the API Console, Claude Enterprise, and Team, but does not support individual plans or third-party platforms.
A tangible incentive. For a protein-design competition, Anthropic and its partners allocated up to $1 million in Claude credits, $250 000 in Modal compute credits, and lab validation for more than 5 000 designs.
For now, Anthropic is keeping LSVP in beta and does not extend it to individual plans or third-party platforms.
