Introducing Claude Fable 5

TL;DR: Anthropic releases its strongest model yet, Claude Fable 5 (Mythos-class), introducing automatic safety review mechanisms alongside advanced capabilities. High-risk requests will be redirected to Opus 4.8 to balance power with safety. ## Model Release and Background Claude Fable 5 is the most powerful model Anthropic has ever released to the public, belonging to the Mythos class. It is equipped with safety measures and is ready for general deployment. Previously, Anthropic did not widely release its previous-generation model of equivalent capability — the Claude Mythos preview. During training and testing, that model discovered thousands of cybersecurity vulnerabilities. A model capable of finding such vulnerabilities could also be used to exploit them. Therefore, Anthropic decided not to release it publicly at that time, instead delivering it to the protectors of critical global software so they could fix the vulnerabilities before they were exploited. The company believes this was the right decision then, but it was never the end goal — because powerful AI should be both safe and accessible. That is why Claude Fable 5 was built. ## Safety Considerations and Design All Claude models have built-in safeguards, but Fable requires more careful safety design than any previous model. Fable 5's safety system automatically reviews requests involving high-risk domains such as cybersecurity or biology, and then redirects those requests to Opus 4.8. This ensures users can benefit from the capabilities of powerful models like Fable while avoiding the associated cybersecurity and biosafety risks. The current scope of protection is broad, and Anthropic will continue to refine it to better identify safe requests. ## Capabilities and Applications Claude Fable 5 is designed to tackle the most ambitious tasks. It can handle more complex problems than any previous model, with a high degree of autonomy — able to run continuously for days without human intervention. Its capabilities extend beyond coding to complex projects in finance, scientific research, economics, law, and other fields that previously required constant supervision. Anthropic recommends that users focus it on truly important problems — the kind that, in hindsight, will make people wonder "why did it take so long to solve?" The company knows the boundaries of Fable 5's capabilities, but what is truly exciting is what users will create with it. --- Source: https://www.youtube.com/watch?v=Y9Wz2PV404E