Anthropic Announces New AI Models with Cost Reductions
Anthropic has introduced its latest AI models, Claude Fable 5.1 and Mythos 5.1, designed to address customer feedback regarding price, data retention, and safeguards. The new Claude Fable 5.1 model boasts improved performance compared to Fable 5, while being approximately 25 percent cheaper on average, and up to 45 percent cheaper for intricate tasks due to cost reductions on previously processed data.
Industry Reactions and Improvements
Notable reactions have come from industry leaders. Dan Shipper of Every has praised Claude Fable 5.1 as an enhanced coding model that is faster, efficient with tokens, and converses naturally. Similarly, Aaron Levie, CEO of Box, noted that during tests, Claude Fable 5.1 was more adept at understanding nuances and ambiguities than its predecessor. On the other hand, Mythos 5.1 performs equally to its predecessor in low reasoning benchmarks, regardless of settings.
Updated Safeguards and Data Privacy
The new models also feature adjustments to safeguards and data privacy. Fable 5.1 possesses “more precise safeguards,” reducing the likelihood of blocking basic biology questions, unlike Mythos 5.1 which retains the same restrictions as the prior version. In terms of data security, Anthropic’s Enterprise Frontier Safeguards will ensure complete data privacy by utilising customer cloud servers for storage, with plans to implement these changes later this fall.
Availability and Specific Use Cases
Claude Fable 5.1 is now available across all platforms, enabling new functionalities such as software vulnerability identification. However, more complex cybersecurity tasks, including penetration testing and exploit generation, will continue to be directed to the Opus models. Meanwhile, Mythos 5.1 remains accessible exclusively to participants in Project Glasswing.
Reporting based on coverage first published by theverge.com.