Anthropic is bringing Fable 5 back to Claude after a nearly three-week break, but the comeback comes with a catch that has annoyed paying users almost as much as the outage itself: the model will be available for a much shorter time and with half the usage people were originally promised.
Fable 5 is set to return on 1 July, while Mythos 5 is already available to a limited set of US organizations. For Claude Pro, Max, Team, and some enterprise customers, free access to Fable 5 runs only until 7 July. After that, users will have to buy ”usage credits” at a much higher price, roughly matching API rates. In other words, the model is back, but the party’s over almost immediately.
What Claude subscribers are getting instead
The bigger irritation is the reduced allotment. Claude plan users will get only 50% of the usage Anthropic originally planned, which makes the revived access feel more like a trial extension than a proper relaunch. That’s a sharp change from the first rollout on 9 June, when Anthropic said subscribers would get two full weeks of unrestricted access to Fable.
That promise collapsed on 12 June, when Anthropic cut off Fable and Mythos after US export restrictions were triggered by an Amazon report warning that the models could be easier to exploit. Anthropic says it moved quickly to address the issue with a stronger safety classifier designed to spot and block the behavior described in that report.
Why Claude users are angry now
Reddit, predictably, has not been impressed. Users say Anthropic is restoring access only to shrink it again, and the criticism lands because the company had already interrupted the original two-week window after just a few days. The math is ugly: some subscribers say they got three days out of 14, and now they are being offered seven days at half strength.
There’s also a product strategy problem here. Anthropic has already faced complaints that Fable was too quick to refuse certain cyber and biology prompts, and that it sometimes dodged sensitive questions instead of answering them cleanly. Telling users that routine coding and debugging will be shifted to Opus 4.8, a less capable model, only adds fuel to the suspicion that the star model is being rationed before it has even settled in.
Anthropic’s safety-first reset may cost trust
Competitors will love this kind of self-inflicted friction. OpenAI and Google have spent years trying to convince developers that premium models can be both powerful and reliably available; Anthropic now looks cautious, but also a little twitchy. In the enterprise AI market, confidence matters almost as much as benchmark wins, and frequent changes to access policies can turn early enthusiasm into churn.
The next question is whether Anthropic treats this as a one-week bridge back to normal access or as a permanent taper. If Fable 5 keeps arriving with strings attached, users will start planning around other models instead of Anthropic’s roadmap. That is not the kind of loyalty any AI vendor wants to train.