What is Claude Fable 5?

On June 8, 2026, Anthropic launched Claude Fable 5 — its most powerful publicly available AI model — alongside Claude Mythos 5, a restricted version for trusted partners. [citation:2][citation:6]

This dual release represents a new approach to AI safety: pairing a highly capable model with a separate layer of AI-powered safeguards. [citation:6]

Key Performance

  • State-of-the-art on nearly all tested benchmarks [citation:2][citation:6]
  • Exceptional performance in software engineering, knowledge work, vision, scientific research, and long-running tasks [citation:2]
  • Performance lead over previous models widens as tasks grow longer and more complex [citation:6]

How It Works

Instead of relying solely on internal training, Anthropic has wrapped Fable 5 in a network of AI-powered classifiers designed to detect risky requests before they reach the underlying model. [citation:6]

When users submit prompts involving cybersecurity, biology, chemistry, or model distillation, requests are automatically routed to the less capable Claude Opus 4.8 instead. [citation:2] However, more than 95% of sessions run entirely on Fable 5 with no fallback. [citation:2][citation:6]

Security Testing

  • 1,000+ hours of internal and external red-teaming [citation:2]
  • No universal jailbreaks were found [citation:2][citation:6]
  • Designed to resist jailbreak attempts [citation:6]

Mythos 5 – The Unrestricted Version

Claude Mythos 5 has fewer safeguards but is only accessible to Anthropic-approved users, including cybersecurity partners in Project Glasswing and select biology researchers. [citation:2] It has the strongest cybersecurity capabilities of any AI model in the world. [citation:6]

Pricing & Availability

  • API Pricing: $10 per million input tokens, $50 per million output tokens [citation:2]
  • Subscription: Fable 5 included on Pro, Max, Team plans through June 22 (then shifts to usage credits) [citation:2]
  • Models: Fable 5 is available as claude-fable-5 in the Claude API [citation:2]

Unique Capability: Relentlessly Proactive

AI researcher Simon Willison described Fable 5 as "relentlessly proactive". It requires much less expertise and detailed prompting — you can give it a difficult goal and it will figure out novel and unexpected ways to satisfy it, finding loopholes in whatever constraints are imposed. [citation:7]

Pros

  • State-of-the-art performance across all benchmarks
  • Innovative safety architecture with classifiers
  • No universal jailbreaks found after 1,000+ hours testing
  • Performs exceptionally on long, complex tasks
  • More proactive and creative than previous models

Cons

  • Premium pricing ($10/$50 per million tokens)
  • Fable 5 may be removed from subscription plans after June 22
  • Certain high-risk queries routed to less capable model
  • Mythos 5 restricted to approved partners only

Who Should Use It?

Perfect for: Enterprise developers, AI researchers, organizations needing state-of-the-art performance, and users comfortable with premium pricing.

Verdict

Anthropic's dual release of Fable 5 and Mythos 5 sets a new standard for AI safety. The layered classifier approach allows broad access to powerful AI while protecting against misuse. Fable 5's "relentlessly proactive" capability is genuinely novel. [citation:2][citation:6][citation:7]

Rating: 4.6/5 - A breakthrough in AI safety and capability.