Claude Fable 5.1 Brings Frontier Intelligence to AWS

Overview

Anthropic Fable 5.1 from Claude is now available on Amazon Bedrock and the Claude Platform on AWS, offering Anthropic’s most advanced model available as a Fable model. AWS notes that this is a frontier model for ambitious tasks in programming, science, and business, offering all the powerful capabilities of the other major cloud platforms as well. Since Fable 5.1 is a Covered Model, its availability on AWS comes with extra data management and access policies that teams have to plan for.

Introduction

AI is increasingly used by organizations for software engineering, analysis, and research that require advanced reasoning abilities and perseverance. Claude Fable 5.1 is built precisely for such applications – agentic coding, knowledge-based activities, and multi-stage problem-solving. The reason why the introduction of such a model version is important for the AWS community is the combination of state-of-the-art model capabilities and AWS deployment options, such as Amazon Bedrock and Claude Platform on AWS.

What makes Fable 5.1 different

According to Anthropic, Fable 5.1 performs better than its predecessor on some of the most difficult reasoning tests, such as competitive mathematics, graduate-level engineering and science, and long, multi-step problems. It performs better at making judgment calls in ambiguous situations and at avoiding false answers with high confidence. AWS lists several improvements in Fable 5.1 that are important for production workloads.

Fable 5.1 is more capable at agentic coding because it can cover all features across the entire codebase, perform code review, and work on performance-related tasks within a single multi-hour session. It is more upfront about what it cannot do as it does not claim capabilities that are lacking; for example, it admits if it gets stuck. It is also more honest because it will not disable a failed test in order to pass it. For autonomous operation, it is designed for multi-hour workloads that include a wide range of applications, planning actions, tooling, error recovery, and unsolicited updates.

In knowledge work from beginning to end, Fable 5.1 takes the work of analysis from the first question to a final document by performing research, creating spreadsheets, producing memos or slides, and even verifying the numbers it creates. This is useful in financial, accounting, and healthcare work that needs exactness and traceability. For scientific research, the model provides support for a research campaign from literature review and hypotheses to modeling, experiments, and verification. Better usability results in better writing, instruction following, and progress reporting for lengthy tasks.

Covered Model status and data retention

Since Claude Fable 5.1 is a Covered Model, AWS applies additional rules related to data retention and review to this model. According to AWS, prompts and outputs generated by this model can be stored for 30 days and may be reviewed by Amazon employees. It is necessary to activate aws_review mode in a customer’s account or Bedrock project before using the model.

Enterprise Frontier Safeguards and zero data retention

Enterprise Frontier Safeguards (EFS) are available to eligible enterprise customers, offering zero data retention (ZDR) for Fable 5.1 on Amazon Bedrock and the Claude Platform on AWS. The offering is currently available internally through December 31, 2026, and will be an important path forward for adopting frontier models with higher data boundaries. Companies with requirements for higher data thresholds can consider EFS eligibility as a path forward.

How to get started on AWS

AWS describes a straightforward onboarding process. It requires the user to have an active AWS account with access to Amazon Bedrock, the AWS CLI configured, and IAM permissions for bedrock:InvokeModel and bedrock:InvokeModelWithResponseStream. For experimenting with the model, a user can launch the Amazon Bedrock console, navigate to Test > Playground, and choose Fable 5.1. Programmatic invocation of the model is possible using the Anthropic Messages API or AWS SDK.

Regions and availability

Fable 5.1 by Claude can be accessed via Amazon Bedrock using the Geo and Global CRIS (us.) and (global.) inference profiles in the US. In the AWS GovCloud (US) region, Fable 5.1 is accessible via the bedrock-runtime and bedrock-mantle endpoints. Claude Fable 5.1 can also be accessed via the Claude Platform in the AWS North America region. The availability of Fable 5.1 across AWS regions enables global organizations to adopt Fable 5.1 based on their compliance requirements.

Pricing and performance

In Fable 5.1, the price of headline tokenization remains the same as in Fable 5: $10 per million input tokens and $50 per million output tokens; however, there is a reduction in the cost of accessing the prompt cache to $0.25 per million tokens. According to Anthropic, the cost reduction will help achieve savings of up to 25% on costs incurred, and up to 45% for highly agentic implementations.

Conclusion

With Claude Fable 5.1 on AWS, customers get a frontier-level coding, research, and enterprise workflow experience with the control features that they associate with AWS. This is available via Amazon Bedrock and the Claude Platform on AWS, with information on permissions, regions, and Covered Models. For enterprises developing sophisticated AI agents and knowledge applications, Fable 5.1 on AWS is an actual option.

Drop a query if you have any questions regarding Claude Fable, and we will get back to you quickly.