Anthropic Releases Claude Fable 5.1 with 75% Prompt Cache Price Cut
Anthropic has released Claude Fable/Mythos 5.1, setting new state-of-the-art benchmark capabilities. The update features a 75% price cut for prompt caching alongside a 70% increase in maximum output token capacity. Drastically lowering caching costs while expanding output length makes executing long-context tasks significantly more economical for developers. This move pushes the broader AI industry toward lower inference pricing and larger context window utilization in production. The 75% prompt cache discount enables developers to reuse lengthy prompt prefixes, such as system prompts or reference material, without paying full computational costs on subsequent requests. Simultaneously, the 70% bump in output tokens allows the model to generate much longer continuous responses like multi-file codebases or detailed reports.
## BACKGROUND
Prompt caching is a technique that stores static context—such as long instructional documents—so language models can skip re-processing redundant prefix data across multiple API calls. Claude is Anthropic's family of large language models widely used for coding, analytical reasoning, and complex document synthesis.