Claude Fable 5 is the best model I have put on real work this year. It is also expensive enough to make you flinch, and it spent the second half of June switched off by a government order. All three of those things are true at once, which is the only honest way to review it.
What it is good at
Hard problems. Fable 5 is where I go when the task is a gnarly bug across several files, a piece of writing that has to be exactly right, or a chain of reasoning where one wrong step ruins the answer. It holds a long thread without losing the plot, it pushes back when I am wrong instead of cheerfully agreeing, and it writes prose that reads like a person wrote it. For high-stakes work, nothing else in my rotation is as reliable.
What it costs
At $10 per million input tokens and $50 per million output, Fable 5 is a premium product with premium pricing. If you throw bulk, low-stakes work at it, you are lighting money on fire. The trick is discipline: use it where being wrong is expensive, and drop to something cheaper for everything else.
Fable 5 is a great chef. You do not hire a great chef to make your morning toast.
The availability problem
For about two weeks in June, Fable 5 was gone, pulled under a US export restriction. It came back on July 1 with new safeguards. That is a real consideration, not a footnote. A model you cannot count on being available is a model you cannot build a critical path around. It works today. Plan for the day it might not.
The verdict
- Best for: difficult code, careful writing, reasoning where mistakes cost real money.
- Skip if: your work is high-volume and low-stakes, or your budget is tight.
- Watch out for: the price, and the reminder that top-tier availability is not guaranteed.
I keep paying for it, and I do not regret it. But I use it like the expensive tool it is, not the default. If you can only afford one premium model and your work is hard, this is the one. If your work is ordinary, you are overpaying, and there are cheaper models in this same set of reviews that will do fine.