Anthropic Releases Claude Fable 5, Debuting Its Restricted Mythos Class

Anthropic has released Claude Fable 5, the first public version of its powerful Mythos AI model. New safeguards are designed to block responses in high-risk areas, falling back to Claude Opus 4.8.

AI & Culture Desk
Updated Nov 5, 2026 2 min read
Anthropic Releases Claude Fable 5, Debuting Its Restricted Mythos Class

Anthropic has launched Claude Fable 5, making its most powerful AI model, previously kept private, accessible to the public for the first time. This release marks the debut of the Mythos class of models to a wider audience, accompanied by significant safety limitations. According to TechCrunch, Fable 5 excels in software engineering, knowledge work, and vision, but incorporates hard safety limits that block responses in high-risk domains such as cybersecurity, biology, chemistry, and distillation. When these limits are triggered, the model falls back to Claude Opus 4.8.

The decision to release a Mythos-class model, which Anthropic had previously deemed too risky for public access due to concerns over capabilities in areas like cybersecurity, is enabled by new safeguards. The Verge reports that these safeguards are designed to block responses in specific high-risk areas. Anthropic stated that in testing, approximately 95% of Fable sessions ran entirely on Fable responses without needing to defer to Opus 4.8. The company also noted that its internal and external red-teaming efforts, including over 1,000 hours of testing for universal jailbreaks, did not yield success.

For organizations requiring access to the advanced model with fewer restrictions, Anthropic is also deploying Mythos 5. This version operates on the same underlying model as Fable 5 but has safeguards lifted in certain areas. Access to Mythos 5 is currently limited to organizations already approved for the Mythos Preview through Anthropic's Project Glasswing, with plans to expand access over time via a trusted-access program. Anthropic did not provide details on how Mythos 5 relates to the preview or the numbering scheme.

The introduction of Fable 5 and Mythos 5 comes with a mandatory 30-day retention policy for all traffic, even for enterprises that previously had zero-retention agreements. Anthropic stated this data will not be used for training but solely to defend against complex and novel attacks, including new jailbreaks, and to identify and reduce false positives. This policy, framed as a safety measure, could establish an industry precedent for mandatory data retention when accessing powerful AI models.

In terms of performance benchmarks, analytics company Hex reported that Fable 5 achieved a 90% score on its core analytics benchmark for complex, long-running analytical tasks, showing strong judgment and attention to nuance. Vibe-coding platform Base44 noted Fable's improvement in "one-shotting full apps" and its excellent tool-calling capabilities. AI workspace platform Genspark found Fable outperformed other models in its evaluations, particularly on tasks like UI design and game coding.

The pricing for both Fable 5 and Mythos 5 is set at $10 per million input tokens and $50 per million output tokens. This represents double the price of Claude Opus 4.8. TechCrunch suggests this higher cost may act as a deterrent to widespread adoption, especially as enterprises scrutinize AI expenses. However, some companies, like shopping rewards platform Rakuten, view the advanced capabilities, such as Fable's ability to reflect on and validate its own work, as essential for enabling highly autonomous operations, potentially justifying the investment.

Sources

  1. 01 Anthropic’s Claude Fable 5 is a version of Mythos the public can access today — TechCrunch — AI
  2. 02 Anthropic releases its first Mythos-class model Claude Fable — The Verge — AI