Anthropic's newest AI models, Fable 5.1 and Mythos 5.1, are up to 45 percent cheaper for complex tasks and less likely to block basic biology questions, according to The Verge. Updates directly address prior customer criticisms regarding price and safeguards for its 2026 models.

Anthropic offers new data safeguards and more cost-effective, secure models to enterprises. However, a mandatory 30-day data retention policy still applies to its core Fable and Mythos models, creating a compromise for users.

While Anthropic makes significant strides to appeal to the enterprise market, companies with stringent zero-retention policies must carefully assess if these new offerings meet their compliance needs.

Enterprise Safeguards vs. Persistent Retention

  • Claude Fable 5.1, Claude Mythos 5.1, Claude Fable 5, and Claude Mythos 5 require 30-day data retention, according to Anthropic's platform policy.
  • Claude Fable 5.1, Claude Mythos 5.1, Claude Fable 5, and Claude Mythos 5 are not available under Zero Data Retention (ZDR) unless expressly authorized by Anthropic, its platform states.
  • The Activity Feed retains data for 6 years, according to Anthropic's platform.

Anthropic's 'Enterprise Frontier Safeguards' are undermined by mandatory data retention. Its most powerful models require 30-day retention, and the Activity Feed retains data for 6 years. A persistent data footprint means true zero-data-retention remains a significant hurdle for many enterprises, suggesting core data retention concerns are unresolved or require special negotiation. Long-term Activity Feed retention also indicates that true data minimization is not yet a core tenet across all Anthropic services, potentially surprising enterprise customers.

Introducing Cheaper, Smarter, and Safer Models

Claude Fable 5.1 is around 25 percent cheaper than Fable 5 and up to 45 percent less expensive for complex agentic tasks, The Verge reports. Significant cost reductions are offered for businesses utilizing Anthropic's advanced AI capabilities.

Fable 5.1 has more precise safeguards, less likely to block basic biology questions compared to Fable 5, according to The Verge. The improvement enhances the model's usability across a broader range of sensitive inquiries. Fable 5.1 is now approved for identifying software vulnerabilities, The Verge notes, expanding its application in cybersecurity.