The model is not the story
Fable 5 is, by any honest measure, extraordinary. Anthropic's new flagship does things that would have sounded like science fiction when I started designing voice interfaces: sustained reasoning over real work, agentic behavior that holds up over hours, a feel in conversation that is genuinely hard to distinguish from a sharp colleague.
And between the codenames and the rumor threads (Mythos, Project Glass, take your pick), it's easy to lose the actual plot. The plot, for me, is not what the model can do. It's how it is being sold.
Fable 5 arrived on June 9 inside the plans people already pay for, with an expiry date attached. Anthropic's own announcement says it is included until June 22. On June 23 it leaves those plans, and using it takes usage credits on top of the subscription. Two weeks of the best model, then a meter.
A subscription with a meter
Until now the deal was simple. You paid a flat fee and got the best the company had, up to a limit. The flagship was the reason to subscribe.
This launch changes the deal. The flagship becomes a separate product, billed by use, and the subscription keeps last year's intelligence. Anthropic says it expects to bring Fable 5 back into the plans when capacity allows. I read it as a trial run for a new business model, announced as a footnote to a model release.
You subscribe for the best model. Two weeks later, the best model has its own meter.
To be clear about what I'm not saying: they are entitled to do it. It is their model, their costs and their price list. Nobody owes me frontier intelligence at a flat fee. I'm equally entitled not to buy it by the token, and I wouldn't.
Why a designer cares
One thing makes this different from other premium tiers. When you buy the cheap photo editor, you know which filters you're missing. When you use the cheaper model, you can't see what it didn't catch. The buyer can't judge the difference by looking at the output.
So the label has to carry it. Which model answered, what the meter stands at, what the next question will cost: those belong on the screen before the click, in plain words. A free market works when the buyer can read what they're buying. Making that readable is our job.
What settles it
Let me be fair about the economics: frontier models cost staggering amounts to train and run, and the companies building them need revenue that scales with capability.
Whether a meter is the right way to get it is not for me to decide, and it shouldn't be for a regulator either. Hayek described competition as a discovery procedure: a way of finding out facts nobody could know in advance. The right price for a new kind of intelligence is one of those facts. If metering the flagship is a mistake, the correction will come from a rival that puts theirs inside the plan.
The alternative people reach for is that everyone should get the same model at the same price. It sounds generous. In practice, when you level, you level down.
In summary
I'll keep using Fable 5, and I'll keep being amazed by it. Being critical of the price list doesn't require being cynical about the craft.
Intelligence, sold separately, is a fair offer. Put it on the label, and let the competition answer.