Anthropic has announced Claude Fable 5.1 and Claude Mythos 5.1, two versions of the same model with different access conditions and safeguards. Fable 5.1 is generally available, while Mythos 5.1 is limited to Anthropic’s trusted-access programmes.
The distinction is aimed at separating broad access from research work with higher consequences if capabilities are misused. Anthropic says Mythos 5.1’s stronger safeguards are designed for cybersecurity and life-sciences research. The company has also established a biology access programme for Mythos 5.1 in partnership with the US government and plans to open registrations for scientists soon.
Fable 5.1 has a defined cybersecurity boundary
Fable 5.1 can be used to discover software vulnerabilities, but Anthropic says it cannot be used to develop exploits. That gives security teams a specific defensive use case while keeping the generally available model outside unrestricted offensive-security work.
The restriction is important because vulnerability discovery and exploit development are closely related activities, but they carry different risks. Anthropic’s stated boundary allows the company to promote the model’s usefulness for identifying weaknesses without presenting public access as permission to turn those findings into operational attacks.
Anthropic also says its new cybersecurity safeguards produce 60% fewer false positives than before. That figure describes the reported behaviour of the guardrails and their handling of false positives; it is not a promise that Fable 5.1 can run security operations autonomously or that every finding will be accurate.
Lower pricing is tied to specific workloads
For typical token-billed workloads, Anthropic estimates that Fable 5.1 costs about 25% less than Fable 5. The company attributes the reduction to lower cache-read pricing. For highly agentic work, Anthropic says the saving can reach about 45%.
Those percentages are estimates from Anthropic and apply to the workload categories it names. The announcement therefore changes the expected cost profile for some users, but the benefit depends on how a team uses the model rather than applying uniformly to every request.
Enterprise privacy will arrive in stages
Anthropic is also presenting Enterprise Frontier Safeguards as offering complete privacy equivalent to a zero-data-retention policy. Under that system, data is stored in cloud infrastructure fully controlled by the customer rather than by Anthropic.
The enterprise offering will roll out in phases from later this autumn. Until it becomes available, eligible customers can use Fable 5.1 with zero data retention. That provides an interim privacy option for organisations that qualify, while leaving Mythos 5.1 restricted to trusted-access programmes.
Anthropic’s announcement changes the access model, pricing estimates and stated guardrail boundary around Fable 5.1. It does not make the public model an exploit-development system, and it does not turn Mythos 5.1 into a broadly available enterprise product. The trade-off is clear: wider access to Fable 5.1 comes with limits on offensive cybersecurity use, while the more tightly controlled Mythos 5.1 path is reserved for selected research contexts.