Claude Fable 5 returned to global availability on July 1, 2026, after a pause tied to export controls. The model had been offline from June 12 to July 1.

On its return, Anthropic set a limited free-inclusion window. Once that threshold is passed, access moves to usage credits, the metered billing mechanism.

The confirmed rate is 10 dollars per million input tokens and 50 dollars per million output tokens. This makes Fable 5 the most expensive model on Anthropic’s current price list.

The free window and its extensions

Anthropic initially included Fable 5 through July 7 for Pro, Max, Team and select Enterprise plans, up to 50% of weekly usage limits. This was a short window, just six days from the model’s return.

The deadline was pushed back twice, however. On July 7 the included access was extended to July 12, and on July 13 it was extended further to July 19, 2026.

As a result, the switch to usage credits at 10 and 50 dollars per million tokens begins on July 20. The thresholds set for each plan still apply; they simply land later.

The model had first launched on June 9 with an included window through June 22. That original deadline never took effect, because the model went dark on June 12 before it arrived.

Differences between plans

The distinction between Enterprise plans is significant. Enterprise Premium seats share the same grace period as Pro, Max and Team, while Standard seats receive no included allowance.

For Standard seats, billing runs through usage credits from day one. If credits are not enabled for the organization, Fable 5 simply will not run.

Anthropic has not published the exact conversion between “50% of weekly usage limits” and the number of tokens, messages or dollars. The source points to the account usage dashboard as the only reliable reference.

The article also distinguishes this change from a separate credit overhaul floated on June 15 for Agent SDK and automation usage, which was ultimately paused before taking effect.

When the premium rate is worth it

At 10 and 50 dollars per million tokens, Fable 5 costs double Opus 4.8 and roughly five times Sonnet 5’s introductory rate. The batch API halves both figures to 5 and 25 dollars, while cache hits drop to 1 dollar per million tokens.

The full one-million-token context window is billed at the same per-token rate. Every current Claude model shares the same one-to-five input-to-output ratio.

According to the analysis, Fable 5 should be treated as a planning model rather than a default for bulk coding. The recommended choice is to reserve it for architecture decisions and the hardest problems.

For production work and repetitive coding, the guidance is to route tasks to Sonnet 5, at roughly a fifth of the per-token cost. Opus 4.8 remains an option for general work, while Haiku 4.5 covers high-volume tasks.

Original article: digitalapplied.com