Claude Fable 5.1 makes cache economics and safety routing part of the deployment decision

Anthropic's September 1 product-page update lists Claude Fable 5.1 for Pro, Max, Team, and Enterprise users, and says developers can access it through the Claude Platform, marketplaces, AWS, Google Cloud, and Microsoft Foundry. The page positions the model for long-running coding and knowledge-work tasks, while the accompanying AWS guidance documents a Bedrock deployment path. Those are provider statements about access and intended use, not an independent measurement of task reliability.

The clearest operational comparison with Fable 5 is pricing for repeated context. Anthropic lists Fable 5.1 at $10 per million input tokens and $50 per million output tokens, with cache reads at $0.25 per million tokens, which it says is 75% below Fable 5. Anthropic estimates that this reduces typical-workload cost by about 25% and highly agentic workload cost by up to about 45%; Axios independently reported the same base prices and cache-read reduction. Those estimates matter most where applications actually reuse large prefixes, rather than for short, single-turn calls.

Access also carries constraints that application teams need to model explicitly. Anthropic says cybersecurity and biology safeguards can route flagged requests to less capable models, and that API customers must configure the fallback behavior. AWS describes Fable 5.1 as a Covered Model: its standard Bedrock path requires aws_review mode, under which prompts and outputs can be retained for up to 30 days for human safety review within AWS. AWS also says eligible Enterprise Frontier Safeguards customers can use zero data retention for internal use through December 31, 2026.

AWS's implementation notes make the integration consequences concrete. Its example uses the global.anthropic.claude-fable-5-1 Bedrock model ID, requires InvokeModel or InvokeModelWithResponseStream permissions, and advises clients to select the text block rather than assume a fixed response position because a reasoning block may precede it. Teams comparing Fable 5.1 with Fable 5 should therefore test their own cache-hit patterns, fallback behavior, retention eligibility, and response parsing. The cited material documents provider claims and deployment mechanics, not a reproducible independent comparison of production quality.