Claude has four models. Think of them as a kitchen brigade.

Why any of this matters to you

You don't need to know which model is answering you, the same way your regulars don't need to know who owns the building. It just works. But sooner or later someone — a vendor, a consultant, your sharpest GM — asks whether you should be on Opus or Sonnet, or what this new "Fable" thing is, and says it like the answer is obvious. It isn't. And most of the people asking couldn't tell you either.

So here's the honest version, in a language you've spoken your whole career. Georges Auguste Escoffier built the modern kitchen a hundred years ago by giving everyone a station and a rank — the brigade. Executive Chef at the top, prep cook at the bottom, and every role earning its place. Anthropic built the same thing and put math-sounding names on it. Peel the names off and it's a brigade. Let me walk you down the line.

The Executive Chef: Fable 5

Fable. This is the one in the news, because it just came back to the whole world on July 1 after a U.S. export-control order got lifted, and it sits at the very top of the lineup.

Think of the Group Executive Chef — the one who doesn't work a station on a Tuesday night. She's the person who can hold your whole operation in her head at once: twelve kitchens, the catering arm, next month's menu rollout, the new unit opening in the spring. You hand her the sprawling, tangled job — "look across everything and tell me what's about to break" — and she directs the rest of the brigade to get it done. Fable is that. It's the most capable model Anthropic ships, it's the one that can read the charts and tables buried inside your PDFs, and it's built for long, multi-step work you hand off and check later instead of babysitting. You don't pull the Executive Chef to dice an onion. You bring her in when the question is big and crosses everything.

The Head Chef: Opus 4.8

Opus. One rung down, and still a heavyweight — the Head Chef who runs a single kitchen at the top of the craft.

This is the one you want on the genuinely hard, deep jobs inside one problem: the complicated buildout, the long chain of reasoning where one wrong step ruins the dish. Opus 4.8 is the most capable model in Anthropic's everyday, generally-available lineup, and it leads on the hardest engineering and the longest agentic work. It's the chef who's been running the pass for fifteen years and doesn't flinch when six tickets land at once. Expensive per hour, worth every penny when the job is hard, overkill when it isn't. Nobody puts their Head Chef on the salad station all night.

The line cook who runs your service: Sonnet 5

Sonnet. Here's the one you'll actually use, and the reason is simple: this is the cook who runs the line every single service.

Sonnet 5 is the new default — the model you get unless you go looking for another — and that's the right call. It's fast, it's a fraction of the cost of the heavyweights, and on most real work it keeps up with them stride for stride. It'll draft the GM email, read last week's sales report, chase the invoice, summarize the lease, answer the ten questions you actually have on a Tuesday. And when a job is big, Sonnet does what a good line cook does — breaks it into pieces and puts a team of prep cooks on it in parallel. Ninety percent of what you'll ever ask Claude to do, this is the station that plates it. A dependable line cook who never calls in sick is worth more than a famous chef you can't afford to run every shift.

The prep cook: Haiku 4.5

Haiku. The bottom of the brigade, and don't you dare read that as "the least important" — try running a Saturday without prep.

Haiku 4.5 is the fastest and cheapest model, built for high-volume, repetitive, good-enough-not-perfect work: sort ten thousand of something into buckets, pull the same three fields off two hundred invoices, label a mountain of stuff, answer the instant back-and-forth where speed beats depth. It's the prep cook dicing a full case of onions before the doors open — you don't need a Head Chef for that, and you'd never pay one to do it. The magic of the prep cook is you can have twenty of them going at once, and they cost almost nothing. Cheap, fast, tireless. Every real kitchen runs on them.

So which one do you actually use?

Mostly, you don't choose — and that's the good news. In Claude Desktop you get the default, which is Sonnet, and for the overwhelming majority of what an operator asks, the line cook is exactly right. The whole reason to know the brigade is to know when to reach past the default.

Here's the rule, plain: for the everyday stuff, take the default and don't overthink it. When a question is genuinely hard and reaches across your whole operation — the kind you'd normally block off an afternoon for — reach up to Opus or Fable, the way you'd call your best chef out for the one dish that has to be perfect. When you've got a mountain of dumb, repetitive sorting, that's prep-cook work; that's Haiku. And if you ever catch a vendor bragging that their product runs on "the most powerful model" for everything — that's the guy who put his Executive Chef on the salad station to look impressive. It's not a flex. It's a tell.

The part the brigade can't fix

One honest catch, because it's the whole game for a multi-unit operator. The best brigade in the world can't cook if the walk-in is chaos — nothing labeled, half of it in the wrong container, the fish where the butter should be. No model, however senior, can tell you which store lost money last Sunday if your sales sit in Toast, your labor in Crunchtime, and your delivery split across DoorDash and Uber Eats, none of it added up anywhere it can reach. Picking the right chef is the easy, fun part. Getting your data prepped — connected, labeled, agent-readable — is the mise en place that lets any of them actually cook for you. That prep is what platforms like Expo exist to do, and it's worth doing no matter which model tops the lineup next year.

Frequently Asked Questions

Which Claude model should I use?

For almost everything, the default — Sonnet 5 — which is what Claude Desktop gives you unless you pick another. Reach up to Opus 4.8 or Fable 5 for genuinely hard, cross-everything jobs, and use Haiku 4.5 for cheap high-volume tasks like sorting or extracting fields at scale.

What's the difference between Opus and Sonnet?

Opus 4.8 is the heavyweight for the hardest work — deep engineering, long multi-step reasoning. Sonnet 5 is the faster, cheaper default that keeps up with Opus on most everyday professional work. Think Head Chef versus the line cook who runs every service: you want the Head Chef on the hardest dish, the line cook on the other ninety percent.

Is Fable 5 better than Opus 4.8?

Fable 5 sits at the top of the lineup as the most capable model, and it came back to global availability on July 1, 2026 after export controls were lifted. Opus 4.8 is the most capable model in the standard everyday tier. Fable is the Executive Chef over the whole operation; Opus is the Head Chef running one kitchen at the top of the craft. Bigger isn't always what you need.

Which Claude model is cheapest?

Haiku 4.5, by a wide margin — built for speed and volume. It's the prep cook: cheap, fast, and ideal for high-volume repetitive work where "good enough" beats "perfect."

Do I have to pick a model in Claude Desktop?

No. You get the sensible default and can just work. Knowing the brigade only helps you decide when a job is worth reaching up for a heavier model — you don't have to think about it for everyday work.