Claude Fable 5.1 and Mythos 5.1 Cut Cache Read Costs by 75%

Anthropic has released updated versions of two of its most capable models, Claude Fable 5.1 and Claude Mythos 5.1, with the headline change being a 75% reduction in cache read costs for Fable. The move is a direct signal to enterprise customers and developers who rely on prompt caching to manage expenses across high-volume applications.

Prompt caching allows repeated content, such as long system prompts or documents, to be stored temporarily so subsequent requests do not re-process the same tokens at full price. For teams running thousands of daily queries against a fixed context, cache read pricing directly affects the economics of deployment. A 75% cut on those reads is a meaningful change for anyone operating at scale. The Fable and Mythos lines have had a turbulent few weeks, having faced temporary disabling following a U.S. government order before being restored to general availability.

What Changed in the 5.1 Releases

Beyond the pricing adjustment, Anthropic has not disclosed sweeping architectural changes for either model in the 5.1 update. The releases appear focused on stability, cost accessibility, and incremental performance refinements rather than a full capability overhaul. Mythos 5.1 receives its own set of updates, though Anthropic's pricing changes center specifically on the Fable tier for cache reads. Developers using Claude's model family across different tiers will want to review updated API documentation to understand exactly where savings apply.

The timing of the release comes shortly after Anthropic restored access to both models following their brief removal. The company debuted Claude Science and restored Fable 5 and Mythos 5 to users in recent weeks, suggesting a period of active iteration across its top-tier offerings. Releasing 5.1 versions so quickly indicates Anthropic is moving fast to refine these models based on real-world developer feedback.

A 75% reduction in cache read costs removes a real friction point for teams building document-heavy or context-intensive applications on top of Fable.Venturebeat

Get the Claude AI Handboek

458 pages on getting more out of Claude, by AI expert Leon Tindemans. A printed book, written in Dutch, shipped worldwide with track and trace.

View the book →

Context and Competitive Pressure

Price reductions on high-capability models have become a recurring theme across the AI industry. Competitors have aggressively cut inference costs over the past year, and Anthropic is responding in kind. Fable represents the company's frontier-level offering in terms of raw capability, and making it cheaper to run cached workloads helps defend its position among developers who might otherwise route traffic to lower-cost alternatives.

The original launch of Claude Fable 5 and Claude Mythos 5 positioned both models as significant steps forward in reasoning and instruction-following. The 5.1 update does not appear to reset those capability claims but instead addresses the commercial side of adoption. For organizations that evaluated the models positively on performance but hesitated on cost, the new cache read pricing may shift that calculation.

It is worth noting that cache read pricing is separate from input and output token costs. Teams should audit their specific usage patterns before projecting savings. Applications that rely heavily on long, repeated system prompts or document contexts will see the largest benefit, while use cases dominated by unique input tokens each request will see less direct impact from this particular change.

Anthropic has been moving its model lineup at a faster clip than many observers expected at the start of the year. With Fable 5.1 and Mythos 5.1 now live, attention will turn to whether further refinements or new model tiers are in the pipeline. Developers can access both updated models through the standard API today.