Fable vs. Opus: My Token-Budget Verdict
AI Coding · 2 min read

Fable vs. Opus: My Token-Budget Verdict

I burned two paid Max accounts down to zero in three days. Let me tell you how.

The Setup

I run two Max accounts, paying monthly — if I remember correctly, $200, because one account alone is not enough. Both on what they call the 20x tier. I bounce between them, or switch during the week, because I’m running multiple projects at the same time.

Normally that setup lasts me the whole week. The model itself tells me when I’m close — usually on a Friday, sometimes a Saturday — “hey, you need to watch your credit, you’re near your weekly budget.” Fine. I get it. Multiple projects, simultaneous work. That’s the deal.

Then I Tried Fable

You probably know Fable already — the most famous, or maybe infamous, model, supposedly built to solve whatever you throw at it. Same story as Spider-Man: with great power comes great responsibility. Except here the great power is Fable’s, and the great responsibility is mine — my own pocket, my own budget. So please, Fable, spend accordingly.

I tried it on Monday. I burned both accounts down almost by Wednesday.

Two accounts. Full weekly budget. Gone in three days instead of a full week. That’s the most infamous token consumer I’ve run into, and it’s shit.

Why I Even Tried It

Because in AI, we all want the thing that’s sure, that doesn’t have issues, that we can rely on. Fable is Fable — everybody’s talking about it. You’ll read LinkedIn posts about people doing amazing stuff with it. Cool for them.

Not for me.

Back to Opus

After I finished burning through my budget, I started reading around, and I thought: holy shit, this isn’t just me. So I switched back to Opus.

And it wasn’t just useful — I’m now using way fewer tokens, and this was the part I didn’t expect: it’s also way faster.

I still think Fable is useful, or at least it seems to be. Maybe you need it if you’re doing something genuinely complex — Fable is supposed to be the best in the world at that kind of complexity. Maybe. I hope so. Who knows.

But for me, and I’d guess for anyone doing SaaS, web apps, that kind of product work — Fable is not needed, at least not right now. I’d strongly suggest switching back to Opus. It works well, it doesn’t burn tokens nearly as fast, and it’s faster — which is the part that matters most when you’re coding every day.

One More Thing: Sonnet for Sub-Agents

While I was at it, I found something else. For spinning up sub-agents and that kind of work, Sonnet is even better. It’s way better at generating text, it’s way cheaper, and — same story — you burn through tokens slower than you’d expect.

So that’s where I landed: Opus for daily coding, Sonnet for sub-agents and text generation, Fable parked until I actually hit a problem complex enough to justify it.

I hope this saves you some time — or at least some money you’d otherwise be sending to Anthropic. Which, by the way, I’m still sending them plenty of. See you in the next one.

← All writing