Claude Fable 5.1 puts migration and data-retention choices in the foreground

Anthropic's current Fable materials and AWS's September 1 guide document Claude Fable 5.1 across native and marketplace paths. AWS lists US Geo and Global Bedrock inference profiles, while Anthropic lists the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry, and Claude Platform on AWS. For teams already operating in several clouds, the immediate change is a wider set of named access routes rather than a single new interface.

The migration notes make the predecessor comparison concrete. Anthropic says three behaviors differ for callers moving from Fable 5: forced tool use errors, earlier models cannot read Fable 5.1 thinking blocks, and editing earlier turns invalidates those blocks. The same documentation lists per-message effort, turn-scoped system messages, progress updates between tool calls, lower cache-read pricing, and content provenance as additions. Existing agent harnesses should test transcript mutation and tool-control assumptions, not just swap the model identifier.

Pricing also changes the tradeoff for long-context agent work. 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 describes as 75% below Fable 5. Axios independently reports the same base input and output prices and the lower cache-read cost. That supports a narrower conclusion than a blanket cost claim: workloads that repeatedly reuse cached context may see a different bill profile, while uncached input and output remain material costs.

The access path has substantive safety and data-handling conditions. AWS says Fable 5.1 requires its aws_review mode, with prompts and outputs retained for up to 30 days for human safety review within the AWS boundary. Anthropic likewise documents 30-day retention by default, while eligible Enterprise Frontier Safeguards customers can use zero data retention. This makes retention mode, eligibility, and review boundaries part of the architecture decision, especially for regulated or document-heavy deployments.

Anthropic's platform documentation draws a second boundary around Mythos 5.1: it says that model shares Fable 5.1 specifications and pricing but is invitation-only through Project Glasswing. Axios separately describes access as limited to vetted partners in cybersecurity and life sciences. The split is an access-policy distinction, not an independent performance comparison, so a technical buyer still needs provider documentation and account-level confirmation for the route it plans to use.