Anthropic has moved Claude Fable 5 out of its standard paid subscription bundle and into metered usage credits for Pro, Max, Team and select Enterprise users, turning its most capable consumer-accessible Claude model into a separate bill on top of a subscription.
Wired's Max Zeff reported in a Wired story that consumers who want to use Claude Fable 5 will need to pay usage-based fees in addition to a Claude subscription, framing the change as the end of the flat-rate AI subscription bargain. Anthropic says the change is tied to capacity. The commercial effect is simpler: the company's highest-end model is no longer treated like an unlimited or near-unlimited feature of the plan.
The timing is important. In Anthropic's June announcement for Claude Fable 5 and Claude Mythos 5, Anthropic said Fable 5 would be included on Pro, Max, Team and seat-based Enterprise plans at no extra cost only through June 22nd, then require usage credits starting June 23rd unless capacity allowed an extension. Anthropic later redeployed Fable 5 with a narrower offer: starting July 1st, Fable 5 was available globally across Claude Platform, Claude.ai, Claude Code and Claude Cowork, but for paid subscription plans it was included only for up to 50% of weekly usage limits through July 7th. After that, Anthropic said, Fable 5 would be available through usage credits.
That makes the current change less a surprise price hike than the close of a promotional access window Anthropic had already described. It still matters because the new boundary puts a price tag on the behavior AI companies have been subsidizing: long-running agentic work, coding sessions, document-heavy analysis and tool use that can consume far more compute than a normal chatbot exchange.
Dario Amodei, Anthropic's co-founder and CEO, has built the company around a familiar tension: ship frontier models broadly enough to compete with OpenAI and Google, while insisting that the most capable systems need safety controls and gated access. Fable 5 is that tension in product form. Anthropic describes it as its most capable generally available Claude model, while Mythos 5, the related higher-capability model, is restricted to approved research and partner programs. Fable 5 carries safeguards that route certain cybersecurity and biology queries to Claude Opus 4.8, and Anthropic says use of Fable requires 30-day data retention for safety monitoring on the Fable product page.
The pricing exposes the compute economics behind that safety-and-access strategy. Anthropic's June post listed Claude Fable 5 and Claude Mythos 5 at $10 per million input tokens and $50 per million output tokens for API use. Its usage credits help page says paid Claude subscribers can enable credits to keep working after hitting plan limits, with additional usage billed at standard API rates and charged separately from the subscription. Anthropic also notes that usage credits apply across Claude conversations and Claude Code terminal usage.
For a consumer product, that is a different promise from the classic $20 or $100 monthly plan. Subscribers still get access to Claude, but the most expensive work can be pushed into a second meter. Anthropic's help center says users can set monthly spending caps, enable auto-reload and prepay for credits. That gives Anthropic a way to avoid hard shutoffs for power users while limiting the losses that come from packing high-cost model calls into a flat subscription. The move also clarifies why frontier AI companies have been careful with model naming and tiering. Anthropic is not simply selling one chatbot. It is segmenting Claude by capability, risk profile, channel and cost. Sonnet-class models can carry much of the everyday workload. Opus sits higher. Fable is being positioned for ambitious, long-running and agentic work, including coding and complex analysis, where Anthropic can justify a per-token meter.
The unanswered question is how temporary the credit requirement really is. Anthropic's June post said it aimed to restore Fable 5 as a standard part of subscription plans when capacity allowed. The July redeployment kept the capacity framing but gave subscribers only a 50% weekly-limit window through July 7th. By July 9th, the practical rule for paid users is metered credits for Fable 5.
That shift will be watched closely by other AI labs because the flat-rate subscription has been one of the cleanest ways to pull consumers and developers into daily AI use. It also creates the worst unit economics when the same users turn chatbots into coding agents, research assistants and workflow runners. Anthropic is drawing the line first around its most expensive consumer-accessible model. If users accept it, the AI subscription business starts looking less like Netflix and more like cloud infrastructure with a membership fee at the door.