L1Lesson 4Free

The model selector: what the options mean in plain English

Pick the right Claude for the job without any technical knowledge

After this, you'll be able to explain what Haiku, Sonnet, and Opus each do in plain terms, check which models your current account can use, and name the kind of task that makes model choice worth paying attention to.

Before you start

Complete Conversation history: what Claude remembers (and what it forgets) first; this lesson builds on knowing how a conversation works by adding the one control that changes how deeply Claude thinks inside it.

The idea

Claude models have different jobs: Haiku is the fast one, Sonnet is the balanced all-rounder, and Opus is the most capable. The exact models and selector you see depend on your plan and surface, so the current account menu is the source of truth.

The learner starts the model selector lesson with stale plan assumptions instead of checking the current account menu.

Sonnet is the dependable choice for writing, questions, summaries, and planning. Haiku is the quick one built for simple, high-volume jobs. Opus is the deep thinker for hard, multi-step, or high-stakes work like a careful contract review.

Here is the before and after: A learner assumes the model name decides whether Claude is useful, and ignores the task. A better learner checks the current selector, keeps balanced work on Sonnet when available, and reaches for Opus before anything delicate, so the careful task gets the careful model.

Now try it open the model control you can see in your current account, if your plan and surface show one. Send a genuinely hard request to your available model, then decide whether the task needs a deeper model or more effort before you trust it.

Use the current selector as the source of truth, then match the model to the task. The point is not the fanciest name; it is enough thinking for the work.

The model selector: what the options mean in plain English mapThe conversation habit works when the setup choice, proof step, and next action stay connected.

Try it (8 min)

Watch out for

Assuming the model name alone decides quality, when the better habit is matching model depth to the task
Trusting old plan notes over the current account menu, which is the only reliable source for what you can choose today
On Pro, reaching for Opus out of habit and spending your usage allowance on trivial tasks Sonnet handles fine
Believing you need to understand the technical specs to choose a model, when the only question is how much thinking the task needs

Paste this into Claude

Open the model control you can see in your current Claude surface, if your account shows one. Then send this hard prompt to the model you have available and read how thoroughly it does:

"Here is a tricky situation: my landlord wants to keep my full security deposit for a small carpet stain that was there when I moved in. I have a move-in photo showing the stain. Walk me through, step by step, how I should respond in writing, what to include, what tone to use, and what my rights likely are. Be thorough."

Notice how much depth your available model gives you on a genuinely hard task.

IF YOUR ACCOUNT SHOWS OPUS: switch to Opus, send the exact same prompt, and note where Opus went deeper.

Finally, tell me in one or two sentences whether your available model covered this for you, or whether the task was hard enough to justify a deeper model when your plan supports it.

What good looks like

You checked the current model control or confirmed your account does not show one
You sent the hard prompt to your default model and read the full answer
You can say in one sentence whether the available model handled the task or needed a deeper model
If your account shows Opus, you tried it and named one concrete way it went deeper

M1 04 Proof PathMove through The model selector: what the options mean in plain, check proof, then fix only the weak part.

When this breaks

Breaks when you treat model choice as a technical decision, because the only practical question is how much depth the task needs, not what the specs are.
Breaks when you rely on stale plan wording instead of the current selector, because model availability changes by plan and surface.

AI can help with this

Wondering whether your task needs a deeper model? Ask Claude itself: 'I want to [describe your task]. Based on the model I have available right now, is this enough, or is it hard enough that I should switch to a deeper model if my plan supports it?'

The lesson rule resolves it by checking the current selector, then matching Haiku, Sonnet, or Opus to task depth.

You can now

You can complete the lesson outcome in a real Claude chat, Project, Artifact, Connector, Desktop, or Code surface.

You checked the current model control or confirmed your account does not show one.
You sent the hard prompt to your default model and read the full answer.
You can say in one sentence whether the available model handled the task or needed a deeper model.
You can verify that if your account shows Opus, you tried it and named one concrete way it went deeper.

Key takeaways

Three model jobs, one question: how much thinking does this need? Haiku is fast, Sonnet is balanced, Opus is deepest. Check the current selector for what your plan can use, then match the model to the task.

Know the three tiers by job: Haiku for fast simple work, Sonnet for almost everything, Opus for hard or high-stakes work.
Use the current account menu and plan page as the source of truth for what you can choose today.
Treat model choice as a depth decision, worth attention when the task is hard enough to need more thinking.
When Opus is available, reserve it for work that is genuinely complex or high stakes.

Go deeper

Was this helpful?

Up nextHow to read Claude's confidence

← Back to Claude Fundamentals

Open the model control you can see in your current Claude surface, if your account shows one. Then send this hard prompt to the model you have available and read how thoroughly it does: "Here is a tricky situation: my landlord wants to keep my full security deposit for a small carpet stain that was there when I moved in. I have a move-in photo showing the stain. Walk me through, step by step, how I should respond in writing, what to include, what tone to use, and what my rights likely are. Be thorough." Notice how much depth your available model gives you on a genuinely hard task. IF YOUR ACCOUNT SHOWS OPUS: switch to Opus, send the exact same prompt, and note where Opus went deeper. Finally, tell me in one or two sentences whether your available model covered this for you, or whether the task was hard enough to justify a deeper model when your plan supports it.