There is a small dropdown at the top of your chat window. It has the model name in it. Most people have never touched it, and a lot do not know it is there.
That dropdown is the difference between the assistant thinking hard about your problem and the assistant answering quickly. Both are useful. They are not the same, and you are probably getting the wrong one for the jobs that matter.
Why the default is not the best one
It is tempting to assume the default is the good one and the rest are lesser options. It is the other way around.
Providers set the default for speed and cost across everybody, not for quality on your particular task. Most messages people send are short and easy, so a fast, cheap, capable model is the sensible thing to hand them. That is a reasonable engineering decision and a bad one for the 4 times a month you need real thinking.
The stronger models are slower and cost the provider more to run, so they are not what you get unless you ask.
As at August 2026: on Claude's Pro plan the strongest available model is Opus 5, but the plan's default is Sonnet 5. So a Pro subscriber who has never opened the dropdown has never used the best model they are paying for. On the Max plan the strongest model is the default.
Model names and tiers change every few months. Do not trust that paragraph in 6 months. Trust the habit instead: open the dropdown, see what is on your plan, and know which one you are talking to.
When it actually matters
Switching for everything is a waste. The fast model is genuinely good, and on simple jobs you will not see a difference worth waiting for.
Use the fast default for: rewording something, summarising a document you pasted in, quick questions with a factual answer, tidying notes, formatting, first drafts of short copy.
Switch to the strongest model for:
- Anything with a decision in it. Pricing, whether to take a job, comparing 2 options where the trade-offs matter.
- Long or messy inputs. A year of numbers, a full contract, 6 months of notes. Holding a lot of context and staying consistent across it is exactly where the stronger model separates.
- Multi-step work. Anything where step 4 depends on getting step 2 right.
- Anything you will act on without checking closely. If you are going to trust it, buy the better version of it.
- When the fast one has already disappointed you. People conclude "AI is not good at this" when the honest answer is that they asked the quick model to do the thinking job.
That last one is the common expensive mistake. The tool got blamed for a setting.
How to check yours, in 2 minutes
- Open a new chat and look at the top of the window for the model name.
- Click it. You will see what your plan gives you access to.
- Note which one is selected by default, because that is what you have been using all along.
- Pick the strongest one and re-run something it did badly last week.
Step 4 is the point. Run a task that disappointed you through the stronger model and compare. That comparison tells you more about what your subscription can do than any amount of reading.
The catch worth knowing
Stronger models usually have tighter usage limits on lower tiers. You get fewer messages before you are asked to wait or switch back.
This is a feature rather than a problem, as long as you are deliberate. Spend the strong model on the 4 things a week that deserve it, and leave the fast one to handle the volume.
The failure to avoid is burning your limit on "reword this paragraph" at 9 in the morning and then hitting the wall at 2, when the pricing decision lands.
The habit
Once a month, open the dropdown and look. Plans change, models get replaced, defaults move, and none of it is announced in a way you will notice.
2 minutes, and it is the highest-return 2 minutes available on a subscription you are already paying for.