Anthropic cuts cache read costs by 75% with Claude Fable 5.1

Anthropic launched two new versions of its top-tier AI models today—Claude Fable 5.1 and Claude Mythos 5.1—just as September kicks off. Both are the same underlying model under different names. Fable 5.1 is the public release with standard safeguards in place. Mythos 5.1 is restricted to vetted cybersecurity and life-sciences teams that need capabilities usually blocked by those safeguards.

The move comes amid intense competition in the AI market, where speed and cost often decide who wins enterprise deals. Anthropic isn’t just boosting performance—it’s cutting the price of cached context reads by 75%. That matters for companies running long-running AI agents that need to recall past interactions without burning through cash. It also introduced Enterprise Frontier Safeguards, a new security framework that lets organizations keep monitoring data inside their own systems instead of sending it to the cloud.

Industry watchers say the price drop is a direct response to customer complaints about the high cost of persistent AI workloads. One early adopter, a cybersecurity firm testing Mythos 5.1, called the reduced cache fees “a game-changer” for running real-time threat detection without breaking the budget. The new safeguards also address concerns about data residency and compliance, letting regulated industries hold onto sensitive logs instead of relying on third-party servers.

Analysts expect other AI labs to follow with similar cost cuts and security tweaks in the coming months. For Anthropic, the timing is strategic—fall is traditionally a big season for enterprise software deals. If companies bite, the changes could help Anthropic gain ground against rivals like OpenAI and Google. The bottom line: cheaper AI agents and tighter control over data could make these new models a must-have for businesses that need both power and privacy.