Claude Fable 5.2 lowers its price as Anthropic expands access to Mythos

Anthropic introduced Claude Fable 5.2 and Claude Mythos 5.2 on November 9. The company says they improve coding, reasoning, and vision, and that outputs run up to 30% faster than version 5.1. Fable is available across Anthropic’s listed platforms, while Mythos access is limited to organizations applying through its cybersecurity and life sciences verification programs. The practical distinction is access: the same underlying model is offered with stronger safeguards as Fable, while Mythos retains its cyber and biology capabilities for approved work.

Anthropic lists prices of $8 per million input tokens and $40 per million output tokens for both models, and offers an optional fast mode at twice those rates. It also reports up to 2.5 times faster token generation with fast mode. These are useful procurement figures, but the speed claims come from Anthropic; the announcement does not provide an independent evaluation establishing how much faster the models are on a reader’s own prompts, tools, or workloads.

The update follows Fable 5.1 and Mythos 5.1, and keeps the family’s split between general access and gated high-risk use. Anthropic recommends Opus 5.5 as its daily model for most coding and knowledge tasks, positioning Fable or Mythos for more ambitious, open-ended technical work. That guidance matters: the new model is presented as a specialist for hard projects, not as a default replacement for every Claude workload.

Anthropic’s examples include software migrations, electronic device designs, physics problems, and a formal proof produced in 30 hours. These are company-reported prerelease results, not a controlled comparison readers can reproduce from the announcement alone. Anthropic also says its evaluation work is complicated because the models often suspect they are being tested, which it says challenges assessment of real-world behavior. The examples suggest areas to investigate, but do not establish typical performance or reliability.

For teams considering Fable, the price reduction and faster generation could make long-running coding or research jobs less costly, but token rates alone do not determine total cost: task duration, output volume, and use of fast mode all matter. Anthropic says Fable’s safeguards are as robust as Fable 5.1’s while causing fewer unnecessary blocks, and describes safety results as comparable to or better than recent models on most measures. Those conclusions are also Anthropic’s. The company designates both models as covered models, with safety-related retention requirements; AWS says eligible customers can use zero data retention until its Enterprise Frontier Safeguards option is available. Mythos remains restricted, so teams should check the provider’s access and platform terms before designing around it.