Claude wrote GPU code and made biology models 4x faster
Open-source biology models are expensive to run, and that cost limits their impact. In its latest Science Blog, Anthropic says Claude optimized inference for more than 30 of them, making them 4x faster on average, partly by writing custom software for GPUs.
Biologists use these models to map molecular structures, design drug-like molecules and predict the effects of genetic mutations. All of the optimization code is going open source on GitHub.
The payoff test is a protein design competition with Adaptyv Bio, where over 5,000 designs will be experimentally validated. Anthropic provides up to $1 million in Claude credits plus funding alongside Adaptyv, with up to $250,000 in compute from Modal and DNA from Twist Bioscience.
Which models gained the most isn't in the announcement; the full results sit in a technical report.
Related stories
- WHO Africa and CEPI turn to Claude in an Ebola outbreak
- Anthropic's new wet lab puts Claude in charge of robots
- Mythos opens to verified biologists for the first time
- Ozempic maker adds Claude next to OpenAI's Rosalind
- Claude does nine loops, one past the N=4 physics record
- Claude found a CRISPR-like puzzle hidden in phage DNA
Comments
No comments yet. Be the first.
Join the conversation
Sign in with Google to leave a comment. Your name and avatar come from your Google profile, and the comment appears after moderation.