Anthropic removed Claude Fable 5 from Pro, Max, Team, and seat-based Enterprise subscription bundles on June 23. Subscribers can still pick Fable 5 inside the Claude app, but each call now draws from prepaid usage credits billed at the API rate of $10 per million input tokens and $50 per million output tokens — double the published price of Claude Opus 4.8. The change ends a 13-day complimentary window that opened when Fable 5 launched publicly on June 9.
Anthropic frames the move as a capacity decision rather than a permanent pricing tier. The company has said demand for Fable 5 was higher than the bundled subscription tiers could absorb, and that it intends to restore Fable 5 to standard plans once infrastructure allows. Usage credits are enabled under Settings > Usage in the Claude app and are billed on the same meter as direct API calls.
Fable 5 has had an unusually eventful first month. The model was briefly suspended in mid-June under a US export-control directive before Anthropic worked through the compliance requirements and restored access. The shift to credit gating now puts Fable 5 in the same access shape as the most expensive tier of the OpenAI and Google frontier stacks: nominally available to all paid users, but metered. Across the three big labs, the binding constraint on the very best models in 2026 is capacity, not subscription tier.
Takeaway for learners: a $50 per million output-token model costs roughly five cents per long answer, which adds up fast if you treat the frontier tier as your default for every prompt. The productive workflow is to iterate cheaply on a smaller model (Sonnet, Haiku, or the Opus tier), cache prompts where the API supports it, and reserve the frontier model for the final pass — the synthesis, the long reasoning chain, the high-stakes review. The cost difference is one or two orders of magnitude, and your output quality almost never benefits by that multiple.