Anthropic has announced the release of Claude Fable 5 and Claude Mythos 5, two new AI models that the company says represent a significant advancement in scientific reasoning, research automation, and life sciences applications. While highlighting the models’ potential to accelerate scientific discovery, Anthropic has also emphasized the need for enhanced safeguards to mitigate risks associated with increasingly capable AI systems.
Claude Fable 5 is being made generally available through Anthropic’s API and subscription offerings. According to the company, the model demonstrates strong performance across a range of scientific and analytical tasks. To address safety concerns, Anthropic has implemented a system in which certain sensitive queries are redirected to its lower-capability model, Claude Opus 4.8.
Claude Mythos 5 is built on the same underlying architecture as Fable 5 but operates with some safeguards relaxed in specific research contexts. Access remains restricted to Project Glasswing participants and selected biology researchers.
Anthropic highlighted several life sciences applications demonstrated during testing and early deployment. Using Mythos 5, researchers reportedly accelerated elements of the drug discovery process by approximately tenfold. The model also matched or exceeded the performance of experienced scientists on selected protein design tasks and independently completed activities including troubleshooting experimental failures, identifying protein binding sites, and selecting and operating protein design tools.
The company reported that Mythos 5 generated promising drug candidates for nine of fourteen protein targets currently undergoing further evaluation. In addition, the model produced novel molecular biology hypotheses and conducted largely autonomous genomics research. Anthropic said the system analyzed millions of cells across 138 animal species and developed a machine learning model that outperformed a recently published model in Science while using substantially fewer parameters.
The launch follows growing attention from government and industry stakeholders regarding the potential risks of highly capable AI systems. Earlier this year, Anthropic introduced Project Glasswing, a cybersecurity initiative designed to help organizations identify and address vulnerabilities in critical infrastructure and widely used software.
In announcing the new models, Anthropic acknowledged that “the same queries that are beneficial in the hands of cybersecurity professionals and biology researchers could be dangerous if available to malicious actors.”
The company noted that existing safeguards for bioweapons-related queries are no longer sufficient given its “reason for concern about well-resourced malicious actors attempting to gain uplift from our models for highly risky biological research.”
Anthropic added that both models demonstrate advanced biological reasoning capabilities, including the completion of a complex gene therapy research task involving adeno-associated viruses (AAVs). While such capabilities could support drug development efforts, the company said they also introduce biosecurity considerations. As a result, many biology and chemistry-related requests submitted to Fable 5 will currently be routed to Claude Opus 4.8 as part of Anthropic’s expanded safety framework.
Click here for the original news story.
