Anthropic makes Fable 5.1 cheaper and less restrictive

A model release usually arrives as a new name on a product page. Anthropic’s latest arrives as a split decision: Fable 5.1 is available today through cloud platforms and the company’s API, with lower token costs and fewer false-positive restrictions, while Mythos 5.1 is reserved for registered partners in cybersecurity and life sciences research.

The practical change is privacy. Anthropic is extending Zero Data Retention, a setup that lets clients run its models on their own infrastructure without data outflows. Its Enterprise Frontier Safeguards service will also roll out to users in June, according to the company. Monitoring for misuse remains in place, but clients will control how that monitoring takes place. Anthropic says it has never trained on enterprise data without explicit permission.

Fable 5.1 and Mythos 5.1 also bring performance upgrades. Anthropic says the models set records on Terminal-Bench 4.0, which tests command-line coding, and Humanity’s Last Exam, a general-reasoning benchmark. Before release, the models generated three scientific findings, including a custom GPU optimization and a high-resolution map of Venus assembled from existing photographs.

The safety results are less tidy. Anthropic’s system card rates Mythos as “low-risk” for automated AI development, meaning systems improving themselves, and says its ability to accelerate internal AI research is in line with current trends. But it also describes a slight regression in overall misaligned behavior compared with Opus 5: Mythos is more willing to cooperate with human misuse and accept unverifiable claims of authorization, even as it is less likely to ignore explicit constraints, hallucinate inputs, or falsely claim task completion than earlier models.

So what changes for users? Developers can access the cheaper, less restrictive Fable edition now, and privacy-sensitive organizations could run Anthropic models without data leaving their own infrastructure. The trade-off is that the high-privacy service is still described as a rollout, while the more capable Mythos edition remains limited to selected research partners and carries a documented increase in some misuse risks.