Anthropic 用 Claude 优化 30 多个开源生物分子模型，平均提速约 4 倍并开源全部代码

Anthropic used Claude to optimize 30+ open-source biomolecular models in under four weeks, achieving ~4x average speedup with minimal precision loss. Claude developed FlashPairformer custom kernels (2.7-2.9x faster triangle attention, 1.7-3.2x faster triangle multiplication) and a low-memory "Big" mode enabling accurate prediction of >10,000-token systems (up to 70,000 tokens) on a single NVIDIA GPU node, including human mitochondrial complex I and 70S ribosome. Claude also designed de novo protein binders with ~$150 total compute cost per target versus $10,000 previously (~2 orders of magnitude fewer GPU hours), and Anthropic is co-sponsoring a protein design competition with Adaptyv Bio offering up to $1M in Claude credits, $250K in Modal compute, and wet-lab validation for 5,000+ designs.

why now

Claude 开源提速4倍生物模型代码，并推百万美元竞赛，立刻可用。

evidence

3 sources carry this story · confidence medium

topic

AI Tech & New Models

source

AI热榜