AI Guides › Getting Started

Reading The Model Picker: What Each Tier Is Actually For

By Nigel Guy · 2 min read

New users tend to read the model picker the way they'd read a phone's storage options — bigger number, better choice, so pick the top one and leave it there. That's an expensive habit in the wrong direction: the biggest, slowest model isn't "the good one" for most of what you'll actually ask day to day, it's the right one for a narrower set of harder tasks.

What the tiers are actually for

Naming and exact tiers change often enough that pinning specifics here would be wrong within months — this is the shape that tends to hold, not a list of current model names to check against a picker that may already look different by the time you read this.

1. A fast, lightweight tier — for quick back-and-forth, simple rewrites, short answers, anything where you'd rather have the reply in two seconds than wait for deeper reasoning you don't need.
2. A capable, balanced tier — the reasonable default for most daily work: drafting, analysis, coding help, most questions that need real thought but aren't unusually hard.
3. The flagship or "thinking" tier — for genuinely difficult reasoning, long or ambiguous tasks, work where getting it right matters more than getting it fast, or where you need it to reason through several steps before answering.

How to actually pick

1. Ask "does this need deep reasoning, or just competent output?" Most everyday writing and quick questions are the second thing.
2. Default to the balanced tier and upgrade only when you notice friction — vague answers, missed nuance, a task that's clearly harder than the tier is handling well.
3. Use the fast tier deliberately for volume work — lots of small, similar tasks where speed matters more than squeezing out the last bit of quality on each one.
4. Reserve the top tier for the task, not the whole day. Switching per task costs nothing; leaving it on the top tier by default mostly costs you usage capacity for no benefit on the easy stuff.

What to skip

Don't leave it on the flagship tier "just in case" for everything — that's solving a problem you don't have yet at the cost of usage you'll want later. Don't downgrade to the fastest tier for something you already suspect is hard just to save a few seconds; that's the one place cutting the corner usually shows.

Guardrails

• Exact model names, capabilities, and what sits in which tier change often — check the current picker and its description in the product itself rather than trusting a remembered mapping.
• Tier isn't the only lever — a clearer question narrows the gap between tiers more than people expect.
• If you genuinely can't tell which tier a task needs, start with the balanced one; it's built to be the reasonable default, not the compromise choice.

All 751 AI guides · JulieMango plans from £17/mo