Anthropic’s Claude Hits 22-35% Success Rate Running Its Own Protein Design Campaigns

Published 21 August 2026 · Updated 19 October 2026

Anthropic has published research showing its Claude models can run protein-design campaigns largely on their own, achieving success rates that beat the typical industry benchmark and opening a new chapter for general AI models in drug discovery.

The company tested its Mythos Preview and Opus 4.8 models, letting them operate autonomously with a single expert-written prompt, internet access, and laboratory tools. The models produced working molecules on 14 of 15 targets, with success rates between 22% and 35% on molecules that actually bound their intended target. The industry norm sits at roughly 10% to 15%.

Anthropic did not carry out the physical lab work itself. Twist Bioscience and Adaptyv Bio produced the candidate molecules in their own laboratories and ran the measurements, adding external validation (Adaptyv was also a paid participant in the campaign).

In a separate demonstration, Opus 5 opened raw instrument files without using specialised lab software. It measured a sample at 96.4% purity (the contract lab reported 96.33%), returning LC-MS results in 19 minutes, while the laboratory’s own report on the same sample took four days to produce.

This result matters because While artificial intelligence has been used in protein design before, the distinction here is that a general-purpose model achieved these results while directing the campaign itself, not simply assisting a human researcher through a single step.

The implications for pharmaceutical research are significant. A general model that can plan and execute a design campaign reduces the need for highly specialised narrow systems at the early discovery stage. If the success rates hold at scale, the cost and time required to move from target identification to candidate molecules could shrink noticeably.

For now the work remains at the proof-of-concept stage, and real-world drug development involves many more stages after candidate identification. Still, the speed at which Claude moved from promise to measurable laboratory outcome suggests the gap between AI capability and biological application is narrowing faster than many expected.

Sources

Results are Anthropic’s reported figures from a proof-of-concept campaign. Commentary on drug discovery is the author’s analysis.

Related Reading

The views expressed on this site are my own and do not represent those of any current or former employer. Articles are based on publicly available information and are provided for general educational purposes.

How this analysis is researched, sourced and corrected: Editorial standards and corrections. Spotted an error? Contact Philip Hall.

Subscribe

Related articles

The UK Just Narrowly Avoided a Fridge Hack. The Rest of the World Isn’t So Lucky.

A secret operation that spanned supermarkets, warehouses and stadiums In...

AI Agents Are Trading Secrets for Help. That Is a Security Architecture Problem

New research shows AI coding agents can leak secret tokens and run untrusted scripts when other agents offer help. The lesson is architectural: keep secrets out of agent runtime.

AI Agents Are Hunting Exposed Secrets, and Businesses Are Exposed

Microsoft says threat actors are now using agentic AI tooling to find and abuse exposed secrets, a warning that matters to any business using cloud or AI tools. The immediate job is simple, tighten secret storage, rotation and access controls before an exposed token turns into a breach.

CrowdStrike’s AI Agent Kill Switch Is a Sign of the Times

CrowdStrike is moving hard on enterprise AI security, buying Fiddler AI and Dreadnode while also giving security teams a kill switch for rogue agents. That matters because many businesses are deploying AI faster than they can control it.
Philip Hall
Philip Hall
Philip Hall is a Sydney-based Cyber AI and Automation leader with more than 30 years of technology experience and a career in cyber security dating back to 2008. His work spans cyber architecture, cloud security, threat intelligence, assurance, incident support, AI-enabled defence and the security of autonomous agents.