Anthropic releases Claude Fable 5, a safety-limited version of Mythos, at $10/$50 per million tokens

Anthropic released Claude Fable 5 on June 9, 2026, marking the first time the company has made its Mythos-class model available to the general public. The model includes hard-coded safety limits that block responses in high-risk domains including cybersecurity, biology, chemistry, and model distillation, automatically falling back to Claude Opus 4.8 for those queries.

Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, double the cost of Opus 4.8. The model is available through Claude's API and consumption-based Enterprise plans.

Access timeline and restrictions

Fable 5 will be included in Pro, Max, Team, and seat-based Enterprise plans at no extra cost through June 22, 2026. Starting June 23, the model will require usage credits, though Anthropic states it plans to restore standard subscription access "as soon as possible."

Anthropic is simultaneously deploying Mythos 5, a more capable version, to organizations already approved for access to the advanced model. Mythos was initially launched as a preview in April 2026 with limited partner access due to cybersecurity concerns.

Safety measures and data retention

According to Anthropic, internal testing including an external bug bounty produced no universal jailbreaks in over 1,000 hours of testing. External red-teaming organizations also failed to find universal jailbreaks. However, the company acknowledges that novel attacks may still be possible.

Anthropic now requires 30-day data retention on all Fable 5 and Mythos 5 traffic, even for enterprises that previously had zero-retention agreements. The company states the data will not be used for training, only to "defend against complex and novel attacks" and "identify and reduce false positives."

Early data shows at least 95% of Fable sessions run entirely on the model's own responses without falling back to Opus 4.8, according to Anthropic.

Performance claims

Analytics company Hex stated that Fable 5 was the first model to achieve 90% on its core analytics benchmark of complex, long-running analytical tasks. Vibe-coding platform Base44 noted improved capabilities in "one-shotting full apps" and tool-calling. AI workspace platform Genspark claimed Fable beat every other model in its evaluations, particularly on UI design and game coding tasks.

Rakuten highlighted the model's ability to "reflect on and validates its own work," stating this enables "highly autonomous operations" where "the extra thinking pays for itself."

What this means

The release establishes a new pattern for deploying frontier AI models: public access with automated safety guardrails rather than restricting access entirely. The mandatory 30-day data retention policy, framed as a safety measure, could become an industry standard as more powerful models are released. The $10/$50 pricing represents a significant cost barrier that may limit widespread adoption, particularly given enterprise concerns about AI spending. Anthropic's approach of offering a capability-restricted public version while maintaining a full-capability version for vetted organizations suggests a two-tier model release strategy may become common for frontier AI systems.