Abstract:
Anthropic announced the launch of Claude Fable 5.1 and Claude Mythos 5.1 on September 1, calling the two models "the world's most advanced programming and knowledge work models." Anthropic says Claude Fable 5.1 sets new standards in programming, knowledge work, and long-term, complex problem-solving tasks.
Under low- or medium-intensity inference settings, this model can achieve equivalent or better results than Fable 5; under higher-intensity inference settings, the performance is significantly improved. Multiple benchmark tests show that Fable 5.1 outperforms Fable 5, Opus 5, and OpenAI’s GPT-5.6 Sol.
Fable 5.1 uses the same basic model as Mythos 5.1, but has different security levels. Fable 5.1 is open to all users, while Mythos 5.1 is currently limited to US enterprises and individual users participating in the Anthropic Trusted Access program. Anthropic also launched the Life Sciences Validation Program, which allows researchers to conduct scientific research using Mythos 5.1’s advanced capabilities in biology, and plans to open registration to scientists soon.
Anthropic said that the security protection mechanism of Fable 5.1 has also been improved. Claude Code users can expect to encounter approximately 60% fewer false positives when handling cybersecurity-related tasks. Anthropic said Fable 5.1 can now be used to find software vulnerabilities, but will not assist in the development of exploits for those vulnerabilities.
In addition, Fable 5.1 adds an invisible watermark to the text generated by Claude. Anthropic is gradually rolling out the corresponding detection API in private preview. In accordance with EU legal requirements, eligible institutions can use this API.
In terms of price, thanks to the reduction in cache read fees, the cost of Fable 5.1 under typical workloads is about 25% lower than that of Fable 5; for highly proxy-based workloads, the cost can be reduced by up to 45%. Base pricing remains unchanged. Fable 5.1 is now live on all platforms.
Comments