Anthropic’s early data shows that more than 95% of Fable sessions involve no fallback at all, and for those sessions, Fable 5’s performance is effectively the same as that of Mythos 5. | Photo Credit: Dado Ruvic

Anthropic on Tuesday released Claude Fable 5, a new Mythos-class model that the AI company has made safe for general and larger use.

Fable 5’s capabilities exceed those of any generally available model, claims Anthropic, adding that it has shown exceptional performance across all benchmarks in software engineering, knowledge work, vision, and other areas. The longer and more complex the task, the larger Fable 5’s lead over our other models, the company said.

Acknowledging the risks that come with stronger models, Anthropic has launched Fable 5 with safeguards that work in a unique way. Instead of shooting down queries or refusing to answer it, Fable 5 will direct select sensitive queries to the next-most-capable model, Claude Opus 4.8.

“Queries on a narrow range of topics, related to cybersecurity, biology and chemistry, and distillation, will instead receive a response from our next-most-capable model, Claude Opus 4.8. Users are informed whenever a fallback occurs, on average in less than 5 per cent of sessions, and performance everywhere else is unaffected. We’ll keep refining the safeguards to reduce false positives,” Anthropic said in its note.

Anthropic’s early data shows that more than 95 per cent of Fable sessions involve no fallback at all, and for those sessions, Fable 5’s performance is effectively the same as that of Mythos 5.

Initial users of Fable 5 say that with earlier models, while one broke a project into multiple model-sized tasks and then put together the results, Fable is now capable of conducting an project end-to-end. As Lovable’s CTO says in a testimonial for Claude Fable 5, it “understands what builders mean, not just what they type. Apps that took a hundred prompts a year ago, is now one-shots.”

Anushree Verma, Senior Director Analyst, Gartner, said that Claude Fable elevates agentic automation, however we would yet ask to prioritise use cases where agentic AI delivers proven value. “Incremental addition will not deliver the right ROI and will also increase the bill. The pricing is almost double Opus 4.8 rates though below GPT 5.5, although they claim to optimize tokens as well,” Verma said.

Jaspreet Bindra, Co-founder & CEO, AI & Beyond, said that the real innovation is not capability alone but the attempt to separate beneficial advanced reasoning from dangerous applications, particularly in cybersecurity and biological research. “While it is topping most effectiveness metrics (though not all), its effectiveness must ultimately be judged through sustained real-world use, independent testing and transparency about failures,” Bindra said.

For a small group of cyber defenders and critical infrastructure providers, Anthropic is also launching Claude Mythos 5: the same underlying model as Fable 5, but with the safeguards lifted in some areas.

Claude Fable 5 is available everywhere starting Tuesday, but Claude Mythos 5 is restricted to Glasswing partners and select biology researchers until the trusted access program is available. Pricing for both models is $10 per million input tokens and $50 per million output tokens, almost the same as GPT 5.5.

Published on June 10, 2026