AWS adds Anthropic's Claude Fable 5.2 and Mythos 5.2 to Amazon Bedrock

What happened

Amazon Web Services announced on 9 November 2026, in its own machine learning blog, that two new Anthropic models are available to its customers: Claude Fable 5.2 and Claude Mythos 5.2. Both can be used in Amazon Bedrock and in Claude Platform on AWS.

AWS describes Fable 5.2 as Anthropic's top general-release model for technical work. Mythos 5.2 is aimed at two high-risk fields and is gated. Read the original AWS post for the full details.

Key details

• Fable 5.2 focus: AWS says it handles ambiguous, multi-step work, checks its own results and can keep running without supervision for hours to days.
• Suggested uses: agentic coding across large multi-repository codebases, long-running agents that coordinate subagents, scientific research, financial analysis, and 3D, hardware and motion design in code.
• Mythos 5.2 access: only through Anthropic's verification programs, for security engineering, red teaming, penetration testing, and life sciences work such as biomolecular design.
• Safety claims: AWS reports Anthropic's alignment testing found Mythos 5.2 hallucinates less than earlier models. Anthropic expects Fable 5.2 to behave similarly. These are company claims.
• Declined requests: a declined cyber security query can be retried on Opus 4.8, and a declined biology query on Opus 5. On Bedrock your code must send the retry. A refusal that hit the prompt cache includes a fallback credit token, so the retry is billed as cache reads.
• Data retention: Anthropic labels Fable 5.2 a Covered Model. By default it requires retention for up to 30 days plus human safety review. On Bedrock, Amazon staff do the review inside the AWS boundary, and you set the aws_review retention mode before calling the model.
• Zero data retention: until Enterprise Frontier Safeguards arrive, eligible customers can use zero data retention for internal use through at least 30 June 2027.
• Model ID: global.anthropic.claude-fable-5-2. You can call it with the Anthropic Messages API, or with the Invoke and Converse APIs.
• Availability: Bedrock offers it via US and global cross-region inference profiles. Claude Platform on AWS offers it in every Region where that service runs.

Why it matters

Teams already on Bedrock get a new Claude tier without changing platforms. The retention rules are the main practical catch, because the default setting involves human review of your data.

AWS also points to a claude-api skill in Claude Code that can update model IDs, thinking settings and effort levels. It is open source.

FAQ

How do I try Claude Fable 5.2 on Amazon Bedrock? Open the Bedrock console, go to Test and then Playground, and pick Fable 5.2. You can also call it from code with the model ID above.

Can anyone use Claude Mythos 5.2? No. AWS says it is offered only through Anthropic's verification programs for cyber security and life sciences work.

Does Fable 5.2 keep my data? By default, yes, for up to 30 days with human safety review. Eligible customers can use zero data retention instead, which AWS says is available for internal use through at least 30 June 2027.