Anthropic Brings Its Most Powerful AI Model to AWS - With Guardrails Built In

Anthropic has launched Claude Fable 5 on Amazon Bedrock and its own Claude Platform on AWS, giving enterprise customers access to what the company calls "Mythos-class" capabilities - top-tier performance in software engineering, complex knowledge work, and visual understanding - without the usual safety trade-offs.

Fable 5 sets new records across nearly every benchmark Anthropic tested. It can run coding and analytical tasks for extended periods without human intervention, self-verify its own outputs, and interpret diagrams, charts, and tables embedded in PDFs and other documents. That makes it a strong fit for finance, legal, analytics, architecture, and gaming, where teams regularly work with dense, visual data.

To keep the model safe for broad use, Anthropic built in safeguards that automatically route high-risk prompts - related to cybersecurity, biology, chemistry, or health - to a less capable model, Opus 4.8. Only a small group of vetted customers can access Claude Mythos 5, the unrestricted version, through a limited preview.

On the technical side, AWS customers can access Fable 5 programmatically via the Anthropic Messages API or the Bedrock Converse and Invoke APIs. Console support is coming soon. Note that you'll need to opt into data sharing using the Data Retention API before invoking the model; Anthropic requires 30-day retention of inputs and outputs for abuse detection. Pricing is tiered: if a harmful prompt gets rerouted to Opus 4.8, you pay Opus rates for the fallback tokens.

Claude Fable 5 is available now in US East (N. Virginia) and Europe (Stockholm), with more regions on the way. If your account doesn't have access yet, it should be enabled soon based on your Bedrock usage - or you can contact AWS Support to expedite.

Source: AWS