Anthropic releases Claude Fable 5, a 'Mythos-class' model with safeguards for public use

Anthropic has released Claude Fable 5, a public version of its previously restricted Mythos model line, with safeguards designed to prevent cybersecurity misuse.

According to Anthropic, Fable 5 is a "Mythos-class" model that "exceeds the capabilities of any other generally available AI model" the company has published. The company claims the model shows "state-of-the-art" performance across software engineering, knowledge work, vision, and scientific research benchmarks, with advantages increasing for longer and more complex tasks.

Safeguard mechanism

Fable 5 includes a novel safeguard system: when the model encounters restricted topics, it automatically switches the conversation to Claude Opus 4.8. Anthropic says it has tuned this "conservatively," meaning it may trigger more frequently than expected.

The original concern with the Mythos line centered on cybersecurity capabilities. The earlier Mythos Preview model could identify and exploit digital vulnerabilities, prompting Anthropic to restrict access.

Mythos 5 for cyberdefenders

Alongside Fable 5, Anthropic announced Claude Mythos 5, which uses the same underlying model but with "safeguards lifted in some areas." This version will be available only to "a small group of cyberdefenders and infrastructure providers."

Mythos 5 will deploy through Project Glasswing in collaboration with the US government. Anthropic claims it has "the strongest cybersecurity capabilities of any model in the world" and plans to expand access through a "broader trusted access program."

According to Anthropic's benchmarks, both Fable 5 and Mythos 5 show gains over Mythos Preview, Claude Opus 4.8, GPT 5.5, and Gemini 3.1 Pro in agentic coding, tool use, and cybersecurity tasks. Specific benchmark scores were not disclosed.

Limited availability window

Fable 5 is available today to Pro, Max, Team, and seat-based Enterprise plan subscribers, but only through June 23. After that date, the model will require usage credits. Anthropic says it will restore standard subscription access "when sufficient capacity allows," but provided no timeline.

Pricing details for the usage credit system and context window size were not disclosed.

What this means

Anthropic is attempting a dual-track approach: releasing powerful AI capabilities publicly while restricting the most dangerous variants to trusted partners. The automatic fallback to Opus 4.8 represents a new technical safeguard mechanism, though its effectiveness depends on classification accuracy. The two-week public availability window suggests significant compute constraints, which may indicate either high demand or expensive inference costs for this model class.