Lesson 1 of 8
Meet Claude's tiers
6 min read
It's all called Claude — so why does one reply land in a blink and another takes its time and goes deep? Because you're not talking to one model. You're choosing a tier.
One Claude, three tiers
Behind the Claude box sits a family of models, not a single brain. Haiku is the light, quick one. Sonnet is the balanced everyday workhorse. Opus is the heavyweight for the hardest problems. Same Claude, different tier — and you choose which one answers.
There is no single best tier — only trade-offs. More capability costs you speed and money; more speed costs you depth.
The other two dials
Two more knobs shape the same trade-off. Thinking lets Claude answer instantly or take extra time to reason — great for a reasoning-heavy problem, wasted on a one-liner. Where you use it — in the app at claude.ai or wired into your own tools through the API — changes how you reach it, not how smart it is.
Match the tier to the job: quick questions to Haiku, everyday work to Sonnet, the hard stuff to Opus. Don't pay Opus's price for a one-line answer.
The exact tier names shift as new versions ship, and a fresh model can leapfrog the last. Don't anchor on a version number — anchor on the question you're asking and how much depth it deserves.
The shape of it
- —Claude is a family of tiers — Haiku (fast), Sonnet (balanced), Opus (deepest).
- —Capability, speed, and cost trade against each other — there's no free lunch.
- —Pick the tier per task; let it think longer only when the task is hard.
You need a careful, well-argued answer to a knotty contract question. Which tier fits best?
Continue in the app
Take the whole Get the most out of Claude course — tracked
Get your personalized path, progress and streaks in the app — this lesson and every next one, in order.