Fable 5 stopped disappearing. Now it lives in Max.

For five weeks, Claude Fable 5 was the most capable model you could not count on. The US government suspended it three days after launch. Anthropic redeployed it on a six-day timer, then extended that timer again, and again, each reprieve a

For five weeks, Claude Fable 5 was the most capable model you could not count on. The US government suspended it three days after launch. Anthropic redeployed it on a six-day timer, then extended that timer again, and again, each reprieve announced hours before the model was set to vanish. Today the countdown stopped. From July 20, Fable 5 is a standing part of every Max and Team Premium plan. The model itself did not change this week. Its availability did, and for anyone trying to actually build on it, availability was the only feature that was ever missing.

Here is what was announced, what it means for each plan, and the lesson underneath it for anyone whose work now depends on a frontier model someone else controls.

The saga, finally closed

The short version of a chaotic five weeks. Fable 5 launched June 9 alongside its more powerful sibling Mythos 5, positioned as the version of Mythos the public could actually access. Three days later a US export-control directive forced Anthropic to pull both, which we covered when it happened in the US ban on Fable 5 and Mythos 5.

The controls lifted at the end of June, and Fable 5 came back on July 1 on a compressed, capped window, the asymmetric return we wrote up here. Then came the extensions: the deadline moved to July 12, then to July 19, each one announced at the last minute. Today's announcement replaces the next cliff with a standing arrangement.

What actually changes on July 20

The announcement splits by plan, and the split is the part worth reading twice.

Max and Team Premium: Fable 5 is included, at up to 50% of your weekly usage limits, at no extra cost. This is the headline. It is not another temporary extension with a date attached. It is the model becoming a standing part of the plan, per Anthropic's own post.

Pro and Team Standard: access continues through prepaid usage credits, at $10 per million input tokens and $50 per million output tokens, with a one-time $100 credit to soften the transition. So the same model is included on the higher tiers and pay-as-you-go on the lower ones.

Mythos 5 is still not part of this

One correction worth making loudly, because the two names get blurred. This is about Fable 5, not Mythos 5.

Fable 5 is the public version. Mythos 5, the stronger of the pair, remains restricted, available only to a set of US organizations under the government's Project Glasswing program. Nothing today changed that. If you were hoping the Max inclusion quietly brought Mythos back to the public, it did not, and the reasons that ban exists have not gone anywhere.

Why it took several tries, honestly

It would be easy to read the serial extensions as indecision. Anthropic's stated reason is more mundane and more believable: capacity. Demand for Fable was hard to predict, so the company rolled it out to subscription plans in stages and extended access as it secured more compute.

That reading holds up. The June disappearance was a policy event, forced from outside. The July extensions were a supply story, a company rationing a scarce, expensive model while it built enough capacity to promise it permanently. Those are two different kinds of unavailability, and conflating them misses what today actually resolves: the supply one. BleepingComputer's framing, that Fable was never permanently leaving, turned out to be the correct read of where this was heading.

The real lesson for anyone building on frontier models

Here is the part that outlasts this news cycle, and it is the thread running through everything we have written about Fable 5.

A model that is free this week, metered next week, and banned the week before is not infrastructure you can plan a business around. Capability is not the scarce resource anymore. Predictable availability is.

— the thing operators should take from five weeks of this

We build AI systems for clients, and the Fable 5 saga is the cleanest recent argument for a principle we already held: do not marry your product to one model's plan, price, or political standing. The teams that had the worst five weeks were the ones who wired a single vanishing model directly into a live workflow. The teams that had a fine five weeks treated the model as a swappable part behind an interface they controlled.

That is the same discipline we apply when we pick a model at all. As we argued in the Sol vs Terra vs Luna breakdown, the right model is a dial you tune against a job and a cost, not a flag you plant. Today Fable 5 got cheaper and steadier for Max subscribers, which is good news. It is good news precisely because the thing that changed was not the model's intelligence. It was whether you can rely on it being there tomorrow, and that, finally, is a yes.