Anthropic rolled out Claude Fable 5.1 to general availability across its API and major cloud platforms, alongside a twin release, Claude Mythos 5.1, restricted to trusted partner programs. Both systems run on the same underlying model. The practical divergence lies in safety controls: Fable 5.1 operates under standard production guardrails for general commercial use, while Mythos 5.1 features specialized safeguard tiers intended for approved defensive cybersecurity and life sciences research.

Cache Cost Reductions and Pricing Economics

Base API token rates remain identical to Fable 5 at $10 per million input tokens and $50 per million output tokens. However, Anthropic cut the price of prompt cache reads by 75% to $0.25 per million tokens wherever usage is billed by token.

Because agentic coding frameworks frequently reload extensive context histories and tool definitions, this change alters operating expenses. Anthropic estimates that typical token-billed workloads will cost roughly 25% less overall than Fable 5, with multi-turn, context-heavy agentic tasks achieving savings up to approximately 45%. Pricing for Claude Mythos 5.1 was not separately broken out in the release announcement.

Divergent Safeguards and Access Paths

Anthropic modified its guardrail precision on Fable 5.1 to permit defensive vulnerability identification, reporting an average of 60% fewer cybersecurity safeguard interventions per Claude Code session relative to Fable 5. Dual-use activities such as binary vulnerability scanning, exploit creation, and penetration testing remain blocked on Fable 5.1 and redirect to Claude Opus models, as do complex life sciences research requests.

Mythos 5.1 handles more permissive workflows for verified participants:

  • Defensive cybersecurity: Mythos 5.1 is restricted to selected US organizations, and Claude Security now uses it. The announcement says the Cyber Verification Program (CVP) currently covers selected Opus and Sonnet models, with Mythos access planned; organizations can register interest through that program.
  • Life sciences: Administered via the Life Sciences Verification Program (LSVP), which has enrolled initial participants for biological research and development.
  • General-purpose access: Fable 5.1 is available via claude-fable-5-1 on the Claude API, Amazon Web Services, Google Cloud, and Microsoft Azure.

Enterprise Frontier Safeguards and API Anti-Distillation

Anthropic also announced Enterprise Frontier Safeguards (EFS), which it says is designed to provide privacy equivalent to zero data retention by keeping data in customer-controlled cloud infrastructure. Anthropic scheduled a phased rollout starting in fall 2026, offering eligible enterprise customers interim zero data retention on Fable 5.1 until EFS deploys.

To deter automated model distillation, Anthropic updated multi-turn API behaviors. Newly created API accounts cannot manually alter prior context in a conversation while retaining the transcript of the model's intermediate thinking traces. Anthropic stated that existing customer accounts remain temporarily unaffected, with broader enforcement planned across future model releases.