Claude Fable 5.1 ships everywhere while Mythos 5.1 stays in two verification programmes
Anthropic released Claude Fable 5.1 on the Claude API and on AWS, Google Cloud and Microsoft Azure, priced at 10 dollars per million input tokens and 50 dollars per million output tokens. Claude Mythos 5.1 goes only to US organisations admitted to the Cyber Verification and Life Sciences Verification programmes. Amazon Bedrock and Claude Platform on AWS offer Fable 5.1 as a Covered Model and hold prompts inside the AWS boundary. The public model and the gated twin now sit on separate access tracks.
Artificial Intelligence··Night
Claude Fable 5.1 opens on the clouds
Anthropic released Claude Fable 5.1 on the Claude API and on AWS, Google Cloud and Microsoft Azure. Amazon opened the same model on Amazon Bedrock and on Claude Platform on AWS, covering the US Geo CRIS and Global CRIS inference profiles plus the bedrock-runtime and bedrock-mantle endpoints inside AWS GovCloud (US). Claude Platform on AWS is offered in North America. Claude Mythos 5.1 goes only to US organisations admitted to the Cyber Verification and Life Sciences Verification programmes. Claude Fable 5.1 sits in general availability while Claude Mythos 5.1 stays inside those two verification programmes.[1], [2]
Anthropic prices cache reads at 0.25 dollars per million tokens
Anthropic prices Fable 5.1 at 10 dollars per million input tokens and 50 dollars per million output tokens, with cache reads at 0.25 dollars per million tokens. The company reports costs about 25 per cent lower than Fable 5 on typical work and up to about 45 per cent lower on agentic tasks. The benchmark figures are Anthropic's own measurements. The company puts Fable 5.1 at 52.6 per cent on Terminal-Bench-Science 0.1 against 24.7 per cent for Fable 5, and at 55.8 per cent on Terminal-Bench 4.0 against 42.0 per cent. Amazon relays that Anthropic reports Fable 5.1 as an improvement over Fable 5 on Anthropic's hardest reasoning tests, covering competition mathematics, graduate-level questions in engineering and the sciences, and long multi-step problems.[1], [2]
Claude Fable 5.1 is held for up to 30 days on AWS
Amazon classifies Claude Fable 5.1 as a Covered Model. Use is subject to data retention for up to 30 days and human review by Amazon personnel, and invoking the model on Amazon Bedrock requires the aws_review mode set before the call. Prompts and outputs stay inside the AWS boundary and do not require sharing with the model provider. Eligible customers can use Fable 5 and Fable 5.1 with zero data retention through 31 December 2026; additional Enterprise Frontier Safeguards later in the year are described as keeping prompts and outputs in the customer's AWS account, under the customer's encryption keys, with automated review and no human review required. Anthropic writes that both Fable versions scored a zero on OSWorld 2.0 where its safeguards intervened, and that its alignment testing has less visibility into very long-context and multi-agent work.[1], [2]
