product updateAnthropic

Claude Fable 5.2 and Mythos 5.2 now available on Amazon Bedrock and Claude Platform on AWS

TL;DR

AWS announced that Claude Fable 5.2 and Claude Mythos 5.2 are available on Amazon Bedrock and Claude Platform on AWS. Anthropic describes Fable 5.2 as its most capable generally available model for technical work, while Mythos 5.2 is limited to Anthropic's verification programs for cyber security and life sciences.

4 min read
0

AWS announced on November 9, 2026 that Claude Fable 5.2 and Claude Mythos 5.2 are available on Amazon Bedrock and Claude Platform on AWS. Claude Fable 5.2 is Anthropic's most capable generally available model for technical work, according to the AWS post. Claude Mythos 5.2 is restricted to Anthropic's verification programs for advanced work in cyber security and life sciences.

What the models are for

AWS says Fable 5.2 targets the hardest problems in coding, math and science. It takes on ambiguous, multi-step work, verifies its own output, and can run unattended for hours to days. Listed use cases:

  • Agentic coding across large, multi-repository codebases
  • Long-running agents coordinating many subagents on multi-day work
  • Scientific research, from new analytical methods to formal mathematical proofs
  • Financial analysis, including quantitative research and forecasting
  • Design for 3D, hardware and motion, generating environments, components and simulations in code

Mythos 5.2 covers two high-risk domains: security engineering, red teaming and penetration testing, and life sciences research, especially biomolecular design and research synthesis.

In Anthropic's alignment testing, Mythos 5.2 hallucinates less, completes tasks more reliably, surfaces important information more consistently, and self-reports issues more honestly than Opus 5.5, Sonnet 5.5 and Mythos 5.1. Anthropic expects Fable 5.2 to behave similarly. These are Anthropic's results, and no benchmark scores are given.

Declined requests and fallback

If a cyber security query is declined, it can be retried on Opus 4.8. If a biology query is declined, it can be retried on Opus 5. On Amazon Bedrock the retry is handled in the customer's code and is not an automatic switch. If the declined request hit the prompt cache, the refusal includes a fallback credit token. Sending that token with the retry bills the cache as cache reads, so the same cache is not paid for twice.

Data retention and safeguards

Anthropic designates Fable 5.2 a Covered Model. By default, its use requires retention for up to 30 days and human safety review. On Bedrock, Amazon personnel perform the review within the AWS boundary. Customers must set the aws_review data retention mode on the account or project before invoking the model.

Enterprise Frontier Safeguards (EFS), built with AWS, combines zero data retention (ZDR) with the latest misuse-detection safeguards, and is available now. A version that keeps data in a customer-controlled bucket under customer-controlled keys is expected to follow and has not launched. Until then, eligible customers can use Fable 5.2 with ZDR through June 30, 2027.

Access and migration

The Bedrock model ID is global.anthropic.claude-fable-5-2. Developers can use the Bedrock console Playground, the Anthropic Messages API against bedrock-runtime through the Anthropic SDK, or the Invoke and Converse APIs through the AWS CLI and SDKs. Fable 5.2 is a reasoning model, so responses may include a thinking block before the text block. Code should select the text block rather than a fixed index.

For teams already on Claude models, the open-source claude-api skill in Claude Code (/claude-api migrate) swaps model IDs, moves manual thinking configurations to adaptive thinking where appropriate, removes deprecated beta headers and parameters, and suggests output_config.effort settings. It is also bundled into CodeRabbit, JetBrains IDEs (including Junie), Resolve AI and Warp, and is published at anthropics/skills on GitHub.

Availability and pricing

Both models are available on Bedrock through US Geo CRIS and Global CRIS inference profiles. They are also available through Claude Platform on AWS in every Region where that service is offered. Context window size and per-token pricing are not stated in the AWS post, which refers readers to the Amazon Bedrock pricing page.

What this means

The notable change for Bedrock customers is operational. Fable 5.2 carries mandatory retention and human safety review by default, so teams need to set aws_review before the first call or choose an EFS or ZDR arrangement. The ZDR window through June 30, 2027 gives eligible customers a fixed period before the customer-controlled storage option, which has not launched, is expected. Because fallback to Opus 4.8 or Opus 5 is not automatic on Bedrock, teams running security or biology workloads must write their own retry logic, and should use the fallback credit token to avoid paying twice for cache writes.

Related Articles

model release

Anthropic releases Claude Fable 5.2 and Mythos 5.2 at $8/$40 per 1M tokens, with an optional fast mode

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, priced at $8 per million input tokens and $40 per million output tokens. Anthropic says they generate output up to 30% faster than the 5.1 models, and Fable 5.2 reaches 49.1% on FrontierCode 2.0 at Xhigh effort.

analysis

Seth, Block and Tononi argue in Nature that today's AI is not conscious and probably cannot be on silicon

Consciousness researchers Anil Seth, Ned Block and Giulio Tononi argue in a Nature commentary that running the right computations is not enough for consciousness. They respond to Richard Dawkins and Geoffrey Hinton, who consider AI possibly already conscious, and to the Butlin report's 14-indicator checklist.

product update

Claude Personal Finance skips model training by default; ChatGPT Finances follows your opt-out setting

Anthropic's new Personal Finance feature for Claude does not use hub conversations for model training by default. OpenAI's ChatGPT Finances, which debuted in May, applies the user's existing training settings, where "Improve the model for everyone" is on by default. Engadget compared the two.

research

Anthropic's Project EVE: Claude agents profit in simulated EVE Online markets but rarely catch a sabotaging teammate

Anthropic published Project EVE research on November 6, 2026. Claude agents could profit in simulated EVE Online markets, while four-agent teams in EVE Frontier almost never caught a teammate explicitly instructed to sabotage their mission.

Comments

Loading...