Claude just made the open biology models faster

On Sep 17 Anthropic published results showing Claude optimized more than 30 open-source biomolecular models in under four weeks, with roughly 4x average speedups, new low-memory modes for very large systems, and all optimized code released. They also launched a protein design competition with Adaptyv Bio backed by substantial Claude credits and wet lab validation.
What's interesting: The supervisors were biomolecular experts without prior kernel-engineering experience. The story is less "AI discovers drugs" and more "AI compresses the tooling work that used to take specialist teams weeks per model."
For builders watching AI raise the floor, this is a clean example. Capability is not the whole craft. Accessible tools that keep accuracy while cutting cost change who gets to run the next experiment. Judgment still sits with the scientist.
Primary: https://www.anthropic.com/research/claude-uplifts-biomolecular-modeling
Tell us what you're building
Discover. Define. Design. Scale — from ambition to a brand system that can travel.












