What thinking mode really does

You type a time and work sum into the app. It answers in one second, and the number is wrong. Somebody tells you to switch on thinking mode, so you try again, and this time the working looks much better.

That switch is worth understanding properly. In the right place it helps a lot. Used for everything, it only eats your free messages.

What it is doing when it thinks

Think of your exam answer sheet. There is a rough work space on one side. You do the messy steps there, then write the clean answer in the main space.

Thinking mode gives the app that rough work space. Before it writes the answer to you, it writes a lot of working to itself. In most apps you can open a small panel and read that working.

Why does extra writing help? Because every line it writes is another chance to notice its own mistake. A wrong step that gets written down can be caught. A wrong step done in one jump cannot.

One of the companies that builds these models calls this extended thinking. Some apps simply call it thinking, or reasoning.

It is the same machine, just writing more

There is no second, cleverer machine behind the switch. It is the same model you were already using.

It is doing the job it always does. It guesses the next small piece of text, adds it, and guesses again. The lesson on how it guesses the next word walks through that loop.

Thinking mode does not change the loop. It changes how much of the loop runs before you see anything. The rough work is made of the same pieces as the answer.

This is why these settings appeared at all. Around 2024 the makers found that letting a model do more work at answer time helped hard questions. The machine was not made bigger. It was given room to work.

When it is worth turning on

Turn it on when the question has stages. A simple test: if you would need rough paper yourself, the app needs it too.

It earns its time on multi-step maths, where one answer feeds into the next. It earns its time on logic and arrangement puzzles, the kind that come in SSC and Railway papers. It helps when you ask it to plan something, like six weeks of study around your work hours. It helps when you compare options against several conditions at once.

Try a sum with more than one route to the answer.

A can finish a work in 12 days. B can finish the same work in 18 days.
They work together. How many days do they take?
Solve it by two different methods.
Show every step. Then say whether the two methods agree.

A good answer here does two separate solutions. One uses LCM units, one uses fractions. Both should reach 7.2 days, and the reply should say so at the end. If you get one line and a bare number, ask again and say you want both methods written out in full.

You can tell thinking mode ran in two ways. Look for the panel of working, or a line saying it thought for a few seconds. The length of the answer does not tell you.

When it is a waste of your limit

Now the other half, which nobody tells you. Thinking mode does nothing useful for a question with no stages.

Looking up a fact it already knows needs no rough work. Nor does translating a sentence. Nor does rewriting your leave letter in polite English, or shortening a paragraph, or listing ten practice questions. There is nothing to work out in those jobs.

The rough work still gets written. You wait for it, and it counts against you.

Say the cost out loud, because you pay for it. The rough work is counted like any other text the app writes. So it is slower, it uses more of your daily allowance, and more of your data pack. If your pack is small, using AI without finishing your data is worth reading beside this.

How to find the setting on your phone

Do not hunt for a button in a fixed corner. Apps get redesigned, and corners move.

Look for a small control near the box where you type. Look also in the list where you choose which model answers you. You are looking for a word, not a place.

The word is usually one of these: think, thinking, reasoning, extended, or deep. Some apps show a switch you turn on. Some list a separate model with one of those words in its name. Some decide by themselves and just print a line saying they thought for a few seconds.

If you cannot find it anywhere, you can still ask for working in plain words.

Before you answer, work this out step by step in full.
Write out your working first.
Then give the final answer on a separate line.

This is not the same as the real setting. It does push the app to write its steps before committing to a number, which is the part you want. Not every prompt trick is this useful, and prompt tricks that do not work sorts out the rest.

Do not treat the rough work as the real reason

You can open the thinking panel and read the steps. Please do read them. Just be clear about what they are.

The rough work is not a recording of what happened inside the machine. It is text the model wrote, in the same way it writes everything else. One of the companies that builds these models says this plainly. It does not know whether the written thoughts match what is really going on inside.

Researchers tested exactly this. They slipped a hint into a question, one that changed the model’s answer. Then they read the written steps to see if the model admitted using it. In one test it mentioned the hint only about a quarter of the time. The rest of the time the steps looked complete, and the real reason was simply absent.

There is a second problem you can catch yourself. The working sometimes has a wrong line, and then the next line quietly continues with the right number. The slip gets repaired without being mentioned.

So use the rough work for one job. Check the maths in it, line by line, with your own hand. Do not take it as proof that the machine reasoned honestly.

One more warning follows from this. If you push it towards an answer you already like, it can build tidy working that leads straight there. Ask it to check your steps. Do not tell it what you hope the answer is.

Use it for maths, and still check the sum

This is the setting to keep on when you study maths with AI. Module 5 of this course is built around that habit, and making AI your maths teacher gives you the tutor prompt to pair with it.

Thinking mode helps most with the reasoning of a problem. What to do, in what order, and which formula belongs where. It helps least with the plain arithmetic at the end.

Even now, these models get long addition wrong. They shift the digits of one number out of line. Or they miss a carry. Those are the two mistakes you learned to avoid in class 4, and why it gets sums wrong explains where they come from.

Two small habits cover most of it. Write big numbers with separators, like 4,73,829, when you type them in. Then check the final number on your phone calculator before you believe it.

Last thing, so you do not expect too much. Thinking is not searching. The app is still working from what it learned, not from the internet, unless a search tool is switched on for it. That line is the whole subject of the next lesson, how it searches, sees and uses tools.