Anthropic is telling Claude subscribers that their newest model is about to cost extra, and the reason is blunt: the company doesn't have enough servers to meet demand.
If you're paying for Claude Pro, Max, or Team and you've gotten used to reaching for Fable 5, mark July 7 on your calendar. That's the day Anthropic stops including the model in your plan's standard usage. After that, using Fable 5 means buying metered credits at the same rate developers pay through the API, $10 per million input tokens and $50 per million output tokens, stacked on top of whatever you already pay for your subscription.
Until then, there's a grace period. Pro, Max, Team, and select Enterprise seat plans can spend up to half of their weekly usage allowance on Fable 5 through July 7, a limit Anthropic has confirmed and that outlets including Bleeping Computer and InfoWorld have reported in detail. Standard Enterprise seats never got that window at all. They've been metered on Fable 5 since it launched.
Anthropic isn't dressing this up as a product decision. Thariq Shihipar, a Claude Code lead engineer at the company, told users directly that pulling Fable 5 out of subscriptions is not Anthropic's long-term plan, and that the model will come back once capacity allows, as Digg reported. No date is attached to that promise. Anthropic itself has described demand for Fable 5 as "very high, and difficult to predict," and that phrase says more about the state of frontier AI infrastructure than any earnings call would.
This isn't even the first time Anthropic has walked back a plan like this under pressure. A separate credit overhaul the company had floated for June 15 got paused after backlash from developers who build on Claude Code, according to coverage from DigitalApplied. The July 7 change is narrower, it only touches Fable 5, not every model, but the pattern is the same: announce a metering change, watch usage patterns, adjust before the deadline if the reaction is loud enough.
This isn't a Claude problem. It's a compute problem, and Anthropic isn't alone in having it. OpenAI's chief financial officer, Sarah Friar, said this year that compute is "a very scarce resource at the moment" and that the shortage would likely stay tight through 2026 and only ease modestly by 2027, according to reporting picked up by BigGo Finance. Anthropic's own published rate limits tell a similar story. At the top developer tier, Sonnet 4.6 allows roughly 4,000 requests per minute and 400,000 input tokens per minute. OpenAI's GPT-5.4 at a comparable tier allows 10,000 requests per minute and 30 million input tokens per minute, a roughly 75-fold gap in input throughput according to an analysis published by DevTk.AI.
That gap explains a lot. Rationing the newest model, rather than throttling every tier Anthropic sells, is how the company protects the base Sonnet and Opus experience while it waits on new data center capacity that reportedly takes 12 to 24 months to come online.
For anyone actually building on Claude, this is the part that costs money, not the part that makes headlines. Startups that wired Fable 5 into coding agents or customer-facing tools now have to decide whether the model is worth paying API rates for on top of a subscription, or whether Sonnet 4.6 and Opus 4.8, both still fully included, are good enough for the job. A flat monthly line item just became a variable one. Anthropic hasn't published credit package sizes or said whether plans will come with any built-in credit allotment, so nobody can model the actual cost yet.
Consumption-based Enterprise customers and anyone hitting the API directly keep full, unmetered access to Fable 5 the whole time. That's worth remembering the next time an AI company talks about a capacity crunch. It's rarely a constraint on revenue. It's a constraint on whichever tier pays the least.
Also read: Zuckerberg Admits Meta's AI Agents Are Not Moving Fast Enough • Zhipu Turns Anthropic's China Tracking Scandal Into a Recruiting Pitch • Amazon is quietly building the AI chips that power your Echo
