Anthropic launches Claude Fable 5.1 and Mythos 5.1: the same model with two levels of safety

Anthropic introduced on September 1, 2026 two new versions of its most advanced model: Claude Fable 5.1 and Claude Mythos 5.1. Technically they are the same underlying model; the difference lies in the safety layers each one ships with.

Fable 5.1 is available to all paid subscribers via the Anthropic API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, and Microsoft Foundry (with the identifier claude-fable-5-1). It's the version with Anthropic's standard production safeguards.

Mythos 5.1 is delivered only through restricted-access programs to vetted cybersecurity and life-sciences organizations. It is the same model but without some of the filters that typically block advanced tasks — designed for teams that need those capabilities for frontier research or defense work.

Performance: both versions scored 52.6% on Terminal-Bench-Science, a demanding scientific reasoning benchmark. Anthropic describes them as "the world's most advanced models for coding and knowledge work".

The biggest commercial change is that cache reads (when you reuse context already sent to the model) now cost 75% less than in the previous version. For applications that repeat a lot of context — like coding agents or chatbots with large knowledge bases — this can translate into significant savings.

Pricing: stays at $10 per million input tokens and $50 per million output tokens, unchanged from Fable 5.

Why it matters: Anthropic confirms the 2026 trend — labs no longer compete only on being smarter, but on making AI more accessible (same model, two filters) and cheaper in real-world use (75% cheaper cache). For a company spending around $5,000 USD per month on the API, the cache discount can mean hundreds of dollars saved monthly without changing anything in their application.