Anthropic announced Claude Fable 5.1 and Claude Mythos 5.1 and says they are the same model with different safeguard levels. Fable 5.1 is generally available for coding and knowledge work. Mythos 5.1 is limited to vetted cybersecurity and life-sciences organizations through trusted-access programs, so its appearance in an announcement does not mean every account can select it. Read Anthropic’s capability, safety, and eligibility details.

Official image for Claude Fable 5.1 与 Mythos 5.1 发布，开放范围不同
Official image from Anthropic; click to open the original source.

Why one model has two release names

Anthropic treats safeguard level as part of the access design. Fable 5.1 is generally offered for coding, professional writing, and long-running tasks. Mythos 5.1 has safeguards configured for cybersecurity and life-sciences work and is limited to participants in trusted-access programs. Users should not assume they can use Mythos, or interpret vulnerability discovery as permission to develop or deploy exploit code.

Fable pricing changes

Anthropic estimates Fable 5.1 will cost about 25% less than Fable 5 on typical token-billed workloads after reducing cache-read prices. For highly agentic workloads, savings may reach about 45%. Savings come from the cache-read price adjustment; the actual amount depends on repeated context, cache hits, and tool usage. Measure real cache rates and cost per completed task rather than treating the largest estimate as a universal discount.

How teams can verify access and performance

  1. Confirm that the current product and plan offer Fable 5.1 and record its default reasoning level. Anthropic says Claude Code defaults to High, while Claude Cowork and Claude.ai default to Medium.
  2. Compare it with Fable 5 on the same codebase, documents, or multi-step workflow. Track quality, time, and cache hits.
  3. For sensitive enterprise material, check eligibility and rollout timing for zero data retention and Enterprise Frontier Safeguards.
  4. For Mythos, confirm that both the organization and its researchers meet the trusted-program requirements and keep use within the approved research scope.

Anthropic says the safeguards reduce false positives in cybersecurity and support vulnerability discovery without extending to exploit development. A capable model can still misjudge a finding, so security professionals need to review its discoveries and proposed fixes.