Anthropic gave Claude subscribers five more days. Fable 5, which was scheduled to leave paid subscription plans on July 7, now stays included through July 12. After that, access requires usage credits billed at API rates. But there is a meaningful new piece of information: Anthropic has publicly committed to returning Fable 5 to subscriptions once capacity allows.

That distinction matters for how enterprise teams should plan.

What Changed Today

On July 8, Anthropic extended the Fable 5 inclusion window by five days. Through July 12, Pro, Max, Team, and select Enterprise subscribers can use Fable 5 for up to 50 percent of their weekly usage limits with no additional cost. The extension followed significant user pushback and a GitHub issue thread that attracted hundreds of upvotes from Claude Code users.

After July 12, the transition is final — for now. Fable 5 exits the subscription bundle and moves to usage credits at:

Model	Input (per M tokens)	Output (per M tokens)
Claude Fable 5	$10	$50
Claude Opus 4.8	$5	$25
Claude Sonnet 5	$2	$10

That places Fable 5 at double the Opus 4.8 rate — the most expensive model Anthropic ships.

This Is Not Permanent

Anthropic has been explicit that removing Fable 5 from subscriptions is a capacity decision, not a permanent pricing change. A Claude Code lead engineer confirmed in the GitHub thread that Fable 5 is expected to return to subscription plans once the company has enough infrastructure to support it.

The mechanism for that capacity expansion is already in motion. Anthropic’s deal with SpaceX to use all available capacity at the Colossus 1 data center adds more than 300 megawatts and over 220,000 NVIDIA GPUs to its compute infrastructure. Early results are visible: Claude Code rate limits have been doubled since the deal, and peak-hour access restrictions on Pro and Max accounts have been removed.

The path from “capacity constrained” to “included in plans” is being actively built.

The Background: Why This Got Complicated

Fable 5 launched on June 9 with a 13-day complimentary window before billing began. That plan fell apart on June 12, when a US government export control directive forced Anthropic to pull the model from global access. It was offline for roughly six days before coming back online on June 18.

Subscribers who wanted to evaluate the model before committing to usage credits had maybe four usable days, not thirteen. The July 7 cutoff arrived before most teams had a fair assessment window.

Anthropic’s extension acknowledges that implicitly. The additional five days are not charity — they are the model access that subscribers were originally promised but did not receive.

What This Means for Business

If you run a team on Claude, the practical decision point is whether to enable usage credits before July 12.

For teams doing high-stakes, low-volume work — executive reports, legal document analysis, complex strategic planning — Fable 5 is worth the premium. The reasoning depth at the frontier level is genuinely different from Opus 4.8 for specific tasks.

For teams running high-volume workflows — content pipelines, automated analysis, customer-facing agents — Fable 5 at $10/$50 per million tokens is not the right tool. Opus 4.8 at $5/$25, or Sonnet 5 at $2/$10 with the Batch API discount, covers the vast majority of production use cases at a fraction of the cost.

The teams that will navigate this well are the ones with actual usage visibility: which models are they running, on what tasks, and at what volumes. If that data does not exist in your organization, the Fable 5 billing transition is a good forcing function to build it.

One practical note for budget-conscious teams: the Batch API applies a 50 percent discount to Fable 5, bringing it to $5/$25 per million — the same as standard Opus 4.8 pricing. For asynchronous pipelines where real-time response is not required, that eliminates the price premium entirely.

The Longer Picture

The on-again, off-again billing situation for Fable 5 reflects a real constraint in how frontier AI gets deployed at scale. Anthropic is managing access to its most capable model against infrastructure that has not caught up with demand, while also navigating government oversight that can interrupt availability without warning.

That volatility is going to be part of working with frontier AI for the foreseeable future. The companies building resilience into their AI stacks — running workflows that can fall back to capable-but-cheaper models, not hard-coding a dependency on any single frontier model — are better positioned than teams that treat access to the absolute best model as a given.

For a deeper walkthrough of tools like this and how they fit together, the free Working With Claude field guide covers the ecosystem end to end. Get the guide.