Remember when OpenAI used to just let you use the latest model until you hit a wall, and then you just waited an hour? Those were simpler times. Anthropic is now taking a much more clinical, almost hostile approach to their usage limits, and it isn’t pretty. Starting July 20, Claude Fable 5 arrives for Max and Team Premium plans, but there is a catch—a big one. You get 50 percent of the regular limits, and those regular limits are simultaneously being slashed by a third.
Do the math on that: you are losing a third of your baseline capacity, and then you only get half of that remainder if you want to use the new model. It’s a double-dip on the restriction. For Pro users, the “welcome” gift is a one-time $100 credit. After that, you are paying API rates. For anyone who actually uses these models for heavy lifting—coding, long-context analysis, or parsing massive PDFs—this is a gut punch. According to The Decoder, this is a sharp reversal from previous signals about how Fable 5 would be integrated into the subscription tiers.
Basically, Anthropic decided that the flat fee model for power users is a losing game. They are essentially telling their most loyal users to stop pretending the chat interface is a productivity tool and start treating it like a metered utility.
It is a blatant cash grab.
This is the classic move of a company that realized its compute costs are eating its margins. But let’s be clear: this is a terrible user experience. Imagine paying for a premium gym membership only to find out that the squat rack now costs an extra five dollars per set. It’s a friction point that ruins the flow of actual work.
The move to API pricing for Pro users is the real story here. API billing is a nightmare for the casual user (and even for some of us). You have to set up a separate billing account, monitor your credits, and live with the constant anxiety that a runaway loop or a massive prompt will drain your wallet while you sleep. It’s the difference between buying a monthly subway pass and having to swipe a credit card every time you step on the train. Why would anyone want to move from a predictable monthly subscription to a variable cost model?
(Unless you’re a CFO who loves spreadsheets).
The logic is simple: high-volume users are expensive. By pushing them toward the API, Anthropic ensures they never lose money on a power user again. But in doing so, they are alienating the very people who provide the best feedback loops for model refinement. If the tool becomes too expensive to use for an entire project, people will just stop using it for entire projects.
Anthropic is betting that the quality of Fable 5 is high enough that we’ll pay whatever they ask. It’s a bold bet, but it ignores the reality of the current market. With open weights models getting better every month, the cost of switching is lower than it’s ever been. The gap between the top-tier proprietary models and the best open ones is narrowing, and the friction of managing an API key is a significant deterrent for someone who just wants to prompt a model and get an answer.
Within 8 weeks, we will see a significant migration of these squeezed Pro users over to OpenRouter or similar aggregators to hedge their bets across multiple models. If you have to pay by the token anyway, you might as well have the option to switch to a different model the second the current one starts hallucinating or gets lazy.
The “all-you-can-eat” era of LLMs is officially dead. We are now in the era of the meter.
