What Claude Fable 5 Is and Why Safety Routing Matters
Claude Fable 5 is a Mythos-class AI model offered for broad Anthropic API access that uses safety routing AI mechanisms to keep powerful capabilities available to standard users while redirecting higher‑risk prompts to safer alternatives. Anthropic positions Fable 5 as its most capable generally available Claude model, matching the underlying Claude Mythos 5 model but with tighter safety controls. The key idea is that capability and control can be separated: Mythos 5 stays behind trusted access programs, while Fable 5 exposes nearly the same power to a wider audience. To do this, Anthropic adds classifiers that detect sensitive cybersecurity, biology, chemistry, and model‑distillation requests and automatically route them to Claude Opus 4.8 instead. This design brings a Mythos-class model into everyday development and knowledge work without exposing all high‑risk behaviors to the public.
How Safety Routing AI Balances Capability and Control
Anthropic’s safety routing AI acts like a traffic controller between user prompts and different Claude models. When a query is flagged as sensitive—for example offensive cyber tasks, biology and chemistry help, or attempts to copy Claude into another model—Fable 5 does not answer directly. Instead, those prompts go to Claude Opus 4.8, which already supports Dynamic Workflows and stricter behaviors. Anthropic reports that “the fallback path triggers, on average, in less than 5% of sessions,” so more than 95% of Fable 5 use cases receive direct Mythos-class responses. The company also notes that external bug bounties and red‑teaming efforts have not produced universal jailbreaks across more than 1,000 hours of testing. For enterprises, this routing model offers a way to tap advanced AI capabilities while maintaining clear technical guardrails around riskier domains.
Democratizing Mythos-Class Power Through Claude Fable 5
With the Claude Fable 5 release, Anthropic brings what it calls a Mythos-class model into general reach for developers and paid Claude subscribers. Fable 5 is available in the Claude application and via Anthropic API access as claude-fable-5, giving teams access to long‑context reasoning, improved memory, and stronger vision features. Early anecdotes highlight the step change: Stripe reported that Fable 5 completed a codebase‑wide migration in a 50‑million‑line Ruby codebase in one day, work it estimated would have taken engineers over two months manually. Anthropic also cites top performance on Cognition’s FrontierCode evaluation and Hebbia’s Finance Benchmark, along with strong trading‑analysis results from IMC. These concrete tasks show that Fable 5 is not only a chat upgrade but a practical engine for software engineering, financial analysis, and scientific research across text and images.
Mythos 5, Project Glasswing, and Trusted Access Layers
While Fable 5 is widely available, Claude Mythos 5 remains under trusted access for higher‑risk work. Mythos 5 runs on the same underlying model but lifts some safeguards in areas Anthropic restricts, such as offensive cybersecurity tasks or sensitive scientific domains. Access begins with existing Claude Mythos Preview users, including partners in Project Glasswing, Anthropic’s program for cyberdefenders and critical infrastructure providers. Anthropic plans to extend Mythos 5 to more cybersecurity organizations and selected biology researchers over time. This layered model—public Fable 5, routed fallbacks to Opus 4.8, and restricted Mythos 5—lets Anthropic widen public availability without treating every use case the same way. For security‑focused teams, it offers a clearer gradient: use Fable 5 for most workloads and adopt Mythos 5 under a higher‑trust relationship when they require fewer guardrails for defense‑oriented investigations and research.
Pricing, Data Retention, and the Enterprise Adoption Equation
Anthropic is pairing wider access with clear cost and data‑handling terms that enterprises must weigh. Pro, Max, Team, and seat‑based Enterprise subscribers can use Fable 5 at no additional cost through June 22, turning the Claude Fable 5 release into a limited‑time trial window before usage‑credit pricing begins. After that, both Fable 5 and Mythos 5 are listed at USD 10 (approx. RM46) per million input tokens and USD 50 (approx. RM230) per million output tokens, which Anthropic says is less than half the price of the previous Claude Mythos Preview. A key caveat is data retention: Fable 5 and Mythos 5 carry 30‑day traffic retention, even for customers that previously had zero‑retention terms. Security‑sensitive buyers will need legal, compliance, and procurement review to decide whether larger context windows and agentic workflows justify both the premium tier and the new retention rule.
