Claude Fable 5.2 and Claude Mythos 5.2 have listed standard API prices of $8 per million input tokens and $40 per million output tokens. Anthropic's announcement, retrieved November 9, 2026, describes these as lower prices than the preceding versions. That is the source retrieval date; the available announcement extract does not establish the release date.
For developers deciding whether to evaluate these models, the main considerations are the fast-mode premium, coding results at named effort settings and access conditions. Anthropic recommends starting with Claude Opus 5.5 for most coding and knowledge tasks, then switching models according to the work's scope and complexity. The results below are Anthropic's reported evaluations. This explanation includes no independent model benchmarks or reproduced cost and speed measurements.
Fast mode doubles the listed token prices
The listed standard rates apply to both models. Optional fast mode costs $16 per million input tokens and $80 per million output tokens. Anthropic says it speeds up token generation by up to 2.5 times. That describes generation speed and does not establish how much faster a whole coding job, including tools and tests, would finish.
Coding results depend on effort settings
In Anthropic's FrontierCode 2.0 evaluation, Fable 5.2 at High effort scored 45.3%, compared with Opus 5.5 at Max effort at 45.5%, for roughly half the cost per task. At Xhigh effort, Fable 5.2 reached 49.1%. FrontierCode measures whether a model's code changes would be merged by maintainers of open-source repositories.
The reported cost comparison applies to that evaluation and those settings. It does not establish that every Fable job costs less than an Opus job.
Access and model roles differ
Anthropic says Fable and Mythos share the same underlying model. It describes Fable as having strong cybersecurity and biology safeguards that allow it to be available to all users. Organizations pursuing beneficial applications in cybersecurity or biomedical research can apply for Mythos access through Anthropic's Life Sciences and Cyber Verification Programs.
The announcement also describes Fable as an Advisor, where Haiku, Sonnet or Opus models consult it on harder problems, or an Orchestrator, where it delegates simpler tasks to smaller models. These are documented roles; their integrations were not tested for this explanation.
What to compare on your own workload
Use Anthropic's Opus 5.5 recommendation as a starting point. If evaluating Fable 5.2 for harder, open-ended technical work, keep the task set fixed and record each model's effort setting. Compare accepted results, total charges and elapsed time, with standard and fast-mode runs recorded separately. This proposed comparison would test whether the published cost and speed advantages carry over to your repositories.