OutYet reporting
Claude Fable 5.1 Cuts Cache Read Prices, Not Every Token Price
Anthropic says cheaper cached-context reads lower costs for some repeated-context workloads. The size of the bill reduction depends on how much of a run uses cached input.
Anthropic’s Claude Fable 5.1 keeps the same headline input and output prices as Fable 5, but cuts the price for reading cached context to $0.25 per million tokens. The Claude Platform documentation describes this as a lower cache-read price alongside changes to the model’s API behavior. That distinction matters: the 75% reduction applies to one billing line, not to every token or every request. Developers comparing model costs should separate cache reads from fresh input, cache creation, and generated output.
Anthropic announced Fable 5.1 on September 1, 2026, and says it is available through its platform and cloud marketplaces. Its model page lists the same $10 per million input tokens and $50 per million output tokens as Fable 5, while cache reads fall from $1 to $0.25 per million tokens. Anthropic estimates that this change makes typical workloads about 25% cheaper and highly agentic workloads up to about 45% cheaper. Those are provider estimates, not independently reported measurements of a representative workload.
The practical difference from Fable 5 is therefore workload dependent. A run that repeatedly reads a large, stable prompt or codebase can put a larger share of its input into cache reads and benefit more. A run dominated by fresh input or generated output has less exposure to the discounted line item, so its total cost will fall by less than 75%. VentureBeat’s coverage also identifies the cache-read reduction as the pricing change while reporting Anthropic’s estimates; it does not supply independent benchmark measurements validating those savings.
For API users, the useful comparison is the actual mix of cache reads, cache writes, fresh input, and output in their own traces, rather than the headline discount alone. The official documentation also flags API behavior changes in Fable 5.1, so a cost comparison should keep request patterns and settings consistent when comparing it with Fable 5. The public sources establish the rate change and Anthropic’s projected savings, but they do not show how much a particular application will save. Subscription usage is not priced per token in the same way, so this rate-card change should not be assumed to translate directly into lower subscription limits or costs.