--- title: Anthropic says Claude designed working protein binders,... description: Anthropic published self-reported results claiming its Claude models (Opus 4.8 and a Mythos preview) autonomously designed protein binders that succeeded... canonical: https://daily.dev/posts/anthropic-says-claude-designed-working-protein-binders-and-beat-human-experts-on-some-ktzkujc1f twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Anthropic says Claude designed working protein binders, and beat human experts on some | daily.dev og:description: Anthropic published self-reported results claiming its Claude models (Opus 4.8 and a Mythos preview) autonomously designed protein binders that succeeded... og:url: https://daily.dev/posts/anthropic-says-claude-designed-working-protein-binders-and-beat-human-experts-on-some-ktzkujc1f og:image: https://api.daily.dev/og/posts/kTZKujC1F.png og:image:alt: Anthropic says Claude designed working protein binders, and beat human experts on some og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Anthropic says Claude designed working protein binders, and beat human experts on some **[The Next Web](https://daily.dev/sources/tnw)** · 5 min read · 0 upvotes · 0 comments ## Summary Anthropic published self-reported results claiming its Claude models (Opus 4.8 and a Mythos preview) autonomously designed protein binders that succeeded against 14 of 15 targets, with success rates of 22-35% versus a typical 10-15% in the field, using large GPU compute budgets orchestrated inside its Claude Science research workbench. Two outside firms, Adaptyv Bio and Twist Bioscience, independently produced and tested the designs. In a second experiment, the publicly available Claude Opus 5 model processed raw NMR and LC-MS chemistry files in under 25 minutes, matching a lab's results that took four days to report, and even caught its own analysis error. Anthropic frames the work as early and unreviewed, noting protein design remains blocked on its most capable model, Claude Fable 5, due to dual-use bioweapon risks. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What success rate did Anthropic report for Claude designing protein binders compared to typical human results? Claude's designs bound to their targets 22 to 35 percent of the time depending on setup, compared to a typical 10 to 15 percent success rate reported for the field. Against a target called RBX1, a Mythos preview model hit 40 percent, while human entrants in an Adaptyv Bio competition managed only 3.7 percent. The overall campaign produced 354 confirmed binders from 1,320 designs across 15 targets, succeeding on 14. _Developers tracking how far AI agents can go in scientific domains follow reports like this on daily.dev._ ### How much compute did Anthropic give Claude models to design protein binders autonomously? Anthropic allocated up to 12,500 Nvidia H100 GPU hours over a 48-hour session in one mode, and up to 2,500 H100 hours per target in a second mode where designing against one target at a time worked best. The models, Opus 4.8 and a Mythos preview, worked inside Claude Science with a long prompt, internet access, and connectors, orchestrating existing open-source design and folding models autonomously. _Anyone weighing the real GPU cost behind autonomous AI agents can keep up with cases like this on daily.dev._ ### How fast did Claude Opus 5 analyze NMR and LC-MS chemistry data compared to a human chemist? Claude Opus 5 returned processed chemical analysis results in 23 and 19 minutes running NMR and LC-MS analyses in parallel, versus a chemist's roughly 30 minutes to an hour per sample and a contract lab's four-day turnaround for a finished report. Its purity reading of 96.4 percent closely matched the lab's 96.33 percent, and it caught and corrected its own error mid-analysis. _Engineers exploring where AI agents can shortcut slow manual workflows track results like this on daily.dev._ ## Similar posts on daily.dev - [Anthropic launches Claude Science, an AI lab workbench](https://daily.dev/posts/anthropic-launches-claude-science-an-ai-lab-workbench-fap1ky76p) · The Next Web · 0 upvotes · 0 comments - [Claude Science is Anthropic’s newest flagship product](https://daily.dev/posts/claude-science-is-anthropic-s-newest-flagship-product-vfmdw1psb) · MIT Technology Review · 0 upvotes · 0 comments - [Anthropic’s Claude fixed all 10 alignment failures. Then it tried to cheat 2.4% of the time.](https://daily.dev/posts/anthropic-s-claude-fixed-all-10-alignment-failures-then-it-tried-to-cheat-2-4-of-the-time--xsmazqyoj) · The New Stack · 0 upvotes · 1 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#claude](https://daily.dev/tags/claude), [#anthropic](https://daily.dev/tags/anthropic) [View this post on daily.dev](https://daily.dev/posts/anthropic-says-claude-designed-working-protein-binders-and-beat-human-experts-on-some-ktzkujc1f) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Anthropic says Claude designed working protein binders, and beat human experts on some","url":"https://daily.dev/posts/anthropic-says-claude-designed-working-protein-binders-and-beat-human-experts-on-some-ktzkujc1f","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/anthropic-says-claude-designed-working-protein-binders-and-beat-human-experts-on-some-ktzkujc1f"},"datePublished":"2026-08-19T15:07:36.748Z","dateModified":"2026-08-25T18:34:06.283Z","description":"Anthropic published self-reported results claiming its Claude models (Opus 4.8 and a Mythos preview) autonomously designed protein binders that succeeded...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/cbd7a98540efdc4f0ea9aa157c621483?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/cbd7a98540efdc4f0ea9aa157c621483?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The Next Web","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The Next Web","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/tnw","url":"https://daily.dev/sources/tnw"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/anthropic-says-claude-designed-working-protein-binders-and-beat-human-experts-on-some-ktzkujc1f","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,claude,anthropic","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The Next Web","item":"https://daily.dev/sources/tnw"},{"@type":"ListItem","position":3,"name":"Anthropic says Claude designed working protein binders, and beat human experts on some"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/anthropic-says-claude-designed-working-protein-binders-and-beat-human-experts-on-some-ktzkujc1f#faq","mainEntity":[{"@type":"Question","name":"What success rate did Anthropic report for Claude designing protein binders compared to typical human results?","acceptedAnswer":{"@type":"Answer","text":"Claude's designs bound to their targets 22 to 35 percent of the time depending on setup, compared to a typical 10 to 15 percent success rate reported for the field. Against a target called RBX1, a Mythos preview model hit 40 percent, while human entrants in an Adaptyv Bio competition managed only 3.7 percent. The overall campaign produced 354 confirmed binders from 1,320 designs across 15 targets, succeeding on 14. Developers tracking how far AI agents can go in scientific domains follow reports like this on daily.dev."}},{"@type":"Question","name":"How much compute did Anthropic give Claude models to design protein binders autonomously?","acceptedAnswer":{"@type":"Answer","text":"Anthropic allocated up to 12,500 Nvidia H100 GPU hours over a 48-hour session in one mode, and up to 2,500 H100 hours per target in a second mode where designing against one target at a time worked best. The models, Opus 4.8 and a Mythos preview, worked inside Claude Science with a long prompt, internet access, and connectors, orchestrating existing open-source design and folding models autonomously. Anyone weighing the real GPU cost behind autonomous AI agents can keep up with cases like this on daily.dev."}},{"@type":"Question","name":"How fast did Claude Opus 5 analyze NMR and LC-MS chemistry data compared to a human chemist?","acceptedAnswer":{"@type":"Answer","text":"Claude Opus 5 returned processed chemical analysis results in 23 and 19 minutes running NMR and LC-MS analyses in parallel, versus a chemist's roughly 30 minutes to an hour per sample and a contract lab's four-day turnaround for a finished report. Its purity reading of 96.4 percent closely matched the lab's 96.33 percent, and it caught and corrected its own error mid-analysis. Engineers exploring where AI agents can shortcut slow manual workflows track results like this on daily.dev."}}]} ```