Anthropic's most powerful public model just got a price tag inside its own subscription plans. As of July 8, 2026, Claude Fable 5 no longer draws from the weekly usage limits that Pro, Max, Team and select Enterprise subscribers pay for each month. Every token now bills through usage credits at pay-as-you-go API rates, layered on top of whatever subscription a customer already holds. July 7 was the final day the Mythos-class model was included at no extra cost, capping a fitful, roughly one-week window of bundled access that opened only after the model returned from a government-ordered suspension.
The switch is small in wording and large in consequence. Fable 5 remains reachable inside claude.ai, Claude Code and Claude Cowork, but reaching it now costs $10 per million input tokens and $50 per million output tokens. Anthropic's other frontier options stay bundled: Claude Sonnet 5 at an introductory $2/$10 through August 31, and Claude Opus 4.8 at $5/$25, both included without extra credit charges. Fable 5, by contrast, is metered separately from the first token.
What Changes on the Invoice
The mechanics are straightforward and, for heavy users, expensive. Before July 8, Pro, Max, Team and premium Enterprise seats could route up to 50% of their weekly usage limit to Fable 5 at no additional charge. That allowance is gone. From Settings to Usage on claude.ai, subscribers now enable and cap credit spend themselves, and every Fable 5 call is billed at the model's standard API list price rather than counted against a plan quota.
Those numbers make Fable 5 the most expensive generally available model Anthropic has ever listed. At $10/$50, it runs exactly double Opus 4.8's $5/$25. Consider a single agentic coding session — a medium multi-file refactor pushing roughly 2 million tokens through the model. Weighted toward output, that session lands near $20 in Fable 5 credits, against about $10 of equivalent work at Opus 4.8 pricing. Run a dozen of those a day across a team and the "included" subscription becomes a rounding error next to the metered bill. Prompt caching softens the input side — Anthropic retains a 90% discount on cached input tokens — but output, where agentic work concentrates its cost, carries the full $50 rate.
The context matters. Fable 5 is Anthropic's first publicly released Mythos-class model, and it only came back online on July 1 after a roughly 19-day export-control suspension. The U.S. Department of Commerce had applied controls on June 12 following a report, credited to Amazon researchers, that a technique could coax the model into identifying and demonstrating software-vulnerability exploits. Anthropic pulled Fable 5 and Mythos 5 for all users rather than try to verify nationality in real time, then shipped a new cybersecurity classifier it says blocks the flagged technique in more than 99% of attempts. With controls lifted on June 30, the model returned — but into a billing regime very different from the one it left.
The Economics Behind the Meter
For Anthropic, usage-credit billing is a capacity valve as much as a revenue lever. Mythos-class inference is costly to serve, and bundling it into flat subscription limits invites exactly the kind of unbounded demand that a freshly restored, capacity-constrained model can least afford. Pricing each token at API list rates makes consumption self-throttling: users who value Fable 5's edge on hard agentic tasks will pay for it, and those who don't will fall back to Opus 4.8 or Sonnet 5, both of which stay inside the plan.
Anthropic has been explicit that the arrangement is temporary. A Claude Code lead engineer said the company does not intend to keep Fable 5 as a permanent paid add-on and expects it to return to standard subscription plans "once capacity allows." That framing positions the credit model as a bridge through a supply crunch rather than a repricing of the flagship — a distinction developers will judge by how long the meter stays on.
For developers, the practical read is a workflow decision, not just a budget line. The included Sonnet 5 and Opus 4.8 tiers now cover the routine majority of tasks, while Fable 5 becomes a deliberate escalation reserved for the problems where its capability gap actually pays for itself. Teams that treated Fable 5 as a default during the free window will feel the reset first, and the Settings-level spend caps become a genuine governance tool rather than an afterthought.
What to Watch Next
Three signals will define the next month. First, whether Anthropic holds to its promise and folds Fable 5 back into subscriptions as capacity recovers, or whether "temporary" quietly hardens into standard pricing for its top tier. Second, how usage credit spend trends now that demand is priced — a sharp drop would confirm the free window was inflating consumption. Third, whether rivals read Anthropic's move as cover to meter their own frontier models, or as an opening to keep theirs bundled. For now, the flagship carries a per-token price inside the plan, and every Fable 5 session is a line item again.
"The company does not intend to keep Fable 5 as a permanent paid add-on and expects it to return to standard subscription plans once capacity allows."— Anthropic, Claude Code team statement