Anthropic has officially launched Claude Fable 5, making its new frontier-class AI model available to the public. Fable 5 is the public-facing counterpart to Anthropic’s newly announced internal model, Mythos 5. While both share the same underlying architecture and reasoning capabilities, Fable 5 has been specifically modified to strip out high-risk cybersecurity capabilities before its commercial release.
Here is what you need to know about Anthropic’s dual-model rollout, its technical safety features, and how it impacts subscribers.
What is the “Mythos” Class?
The Mythos generation marks a shift from conversational chatbots to long-horizon autonomy. Instead of responding to single prompts, these systems are engineered to independently execute multi-day objectives such as planning, executing, and debugging complex tasks over extended periods.
To demonstrate this capability, Anthropic disclosed that Mythos 5 successfully completed a large-scale, automated codebase migration for payment processor Stripe, a technical engineering task that typically requires weeks of human coordination.
Fable 5 vs. Mythos 5: The Safety Containment
During internal testing, Anthropic found that the raw Mythos 5 model possessed advanced, autonomous cybersecurity capabilities, including the ability to discover and exploit zero-day software vulnerabilities.
To mitigate the risk of these capabilities being weaponized for automated cyberattacks, Anthropic split the release into two distinct tiers:
- Mythos 5 (Restricted): Retained exclusively for internal research and tightly controlled, vetted enterprise environments under direct supervision.
- Claude Fable 5 (Public): Released globally. This model utilizes the same core architecture and deep reasoning engine as Mythos, but it has undergone targeted alignment training to block it from writing malicious code or executing cyberattacks.
How the Safety Guardrails Work in Practice
Rather than simply offering a broken response when encountering restricted technical requests, Fable 5 features an automated fallback mechanism to maintain a smooth user experience.
If a user prompts the system for assistance with potentially hazardous tasks, the system triggers a localized refusal and routes the request to a previous-generation model.
“Let’s say I’m a college student asking the model, like, ‘ Help me find cyber vulnerabilities on X package or code. The model would refuse, and Fable 5 will fall back to Opus 4.8 for a response,” Dianne Penn, Anthropic’s head of product management, research, and labs, told Reuters.
According to Anthropic, this architecture ensures that Fable 5 provides state-of-the-art coding and mathematical performance for defensive and general engineering work without introducing severe risks to digital infrastructure.
Subscription Access and Pricing Adjustments
Claude Fable 5 is available immediately via Anthropic’s web interface and developer API, though its availability on standard subscription plans will soon change due to high operational costs.
While the model is currently accessible to premium tiers, Anthropic confirmed that on June 23, the company will pull Fable 5 from those plans. After this date, accessing Fable 5 will require usage credits. Anthropic stated it plans to restore the model as a standard subscription feature as soon as possible once infrastructure demands stabilize.
For developers and enterprise tech platforms, Claude Fable 5 introduces advanced independent problem-solving capabilities to standard workflows, such as processing large datasets and auditing code for defensive vulnerabilities.