Hot City, Hot Takes: Fable Is Back — and the Price Is the Real Story

I'm out on my balcony, watching the city shimmer on one of those hot summer days where even the traffic seems to be moving in slow motion, and I got to thinking about what's actually dominating the tech conversation right now.

Without a doubt, it's the reappearance of Fable, Anthropic's Mythos-class model — its most powerful tier, back online since July 1 after a brief pause tied to export rules. And close behind it, the topic everyone keeps circling back to: the price.

Here's the part worth understanding, even if you're not an engineer.

These AI models charge by the "token" — roughly a chunk of a word, counted both for what you send in and what the model sends back. Fable, the top tier, runs about $10 per million words-in and $50 per million words-out. That sounds abstract until you realize a busy team can burn through millions of tokens in a single afternoon. The sticker shock is real.

But here's the thing: the price tag isn't a wall. It's a dial. Three moves put you back in control.

1. Don't send a Ferrari to pick up groceries

Fable is the premium engine. You don't need it for everything. Anthropic's mid and entry tiers — Opus at roughly half Fable's price, and Haiku at about one-tenth — handle the everyday work just fine: drafting, summarizing, formatting, routine questions. Save the top tier for the genuinely hard problems where a better answer actually changes the outcome. Matching the task to the right tier alone can cut a bill by 5 to 10 times.

2. Stop carrying dead weight

You pay for every token you include, on every single exchange. Paste a giant document into a conversation and then keep chatting, and you're re-paying for that whole document each time you hit send. The fix is simple: include only what the task needs, and start a fresh conversation instead of dragging a bloated one along. Leaner inputs are cheaper — and usually get you sharper answers, too.

3. Reuse and batch

If you keep sending the same background material over and over, "caching" lets you pay for it once and then reuse it at roughly a 90% discount. And if the work isn't urgent — reports that can run overnight, big batches that can wait — processing it in bulk typically costs 50% less than getting it in real time.

The takeaway

Fable being back is exciting, and yes, it's expensive. But treat the cost like a budget you actively steer rather than a meter you watch in horror, and you get the frontier without the panic.

Now — back to watching the city bake.

Want help matching the right AI model to the right job — without the runaway bill? Let's talk about it.