Anthropic speeds biomolecular models, opens design contest

Anthropic said on September 17 that Claude optimized more than 30 open-source biomolecular models in under four weeks, speeding them up about 4x on average and adding a low-memory Big mode that can predict systems larger than 10,000 tokens on a single NVIDIA GPU node. The company is open-sourcing the optimized code, including FlashPairformer kernels for triangle attention, and said a single Claude model with one H200 can now match earlier binder-design campaigns that used about 100 times as many GPU hours. It is co-sponsoring a protein-design competition with Adaptyv Bio, backed by up to $1 million in Claude credits, $250,000 in Modal compute credits, Twist Bioscience DNA, and wet-lab validation for more than 5,000 designs.