Choose the Right AI Brain

Introduction

30 Second Summary

Every question you ask ChatGPT gets the same quick-thinking treatment by default. But it actually has adjustable thinking speeds and specialized tools that can transform the quality of what you get back.

In this project, you will send the same prompt to ChatGPT at different reasoning levels and compare the results side by side. Then you will explore its built-in tools and build a personal AI Decision Playbook that maps your most common tasks to the right capability.

What You'll Build

Picture facing any task and knowing exactly which ChatGPT setting to reach for, getting better results in less time without wasting your usage limits.

By the end of this project, you'll have:

• A side-by-side comparison proving that reasoning level changes output quality more than you expected.
• Firsthand experience triggering Search, Data Analysis, and Image Generation so you know what each tool does and when to reach for it.
• A personal AI Decision Playbook that maps your 10 most common tasks to the right reasoning level and tool combination.
• Secret Mission: Run a Deep Research task and compare its output to what a regular chat produces on the same question.

Are there any prerequisites?

This is Project 3 in the ChatGPT curriculum, so you should have completed Project 1 (Personalize Your ChatGPT) and Project 2 (Stop Prompting Like a Beginner) first.

A ChatGPT Plus subscription is recommended for the full reasoning slider, but Free plan users can follow along using Think mode.

Before We Start

In Projects 1 and 2, you taught ChatGPT about yourself and learned how to write stronger prompts. Now you will discover that the same perfect prompt can produce wildly different results depending on which ChatGPT capability you aim it at.

Find the Reasoning Slider

Every time you send a message, ChatGPT decides how much reasoning effort to put into the response. Sometimes it fires back a quick answer. Other times it pauses to think through something complex.

You have been using ChatGPT this way without knowing you can control that thinking effort yourself. The reasoning slider lets you dial it up or down for every single response. In this step, you will find it, configure a key setting, and identify exactly which thinking levels your plan gives you.

In this step, get ready to:

• Locate the reasoning slider in the model picker.
• Turn off the Higher intelligence auto-switching setting for manual control.
• Identify which reasoning levels are available on your plan.

Locate the model picker and reasoning slider

The model picker is the control panel that lets you choose how ChatGPT responds. It lives right where you type your messages.

• In ChatGPT, start a new chat by clicking New chat in the sidebar.
• Look at the message composer area at the bottom of the screen.
• Find the model picker inside the message composer.

Before you click the model picker, take a guess. What options do you think you will find inside?

• Click the model picker to open it.

You will see a reasoning slider with options like Instant, Medium, and High. This is the control that lets you tell ChatGPT how hard to think about your message.

What is the reasoning slider?

The reasoning slider controls how much thinking effort ChatGPT puts into each response. Instant gives you fast responses for everyday work. Medium provides standard reasoning for more deliberate analysis. High uses extended reasoning for complex multi-step work.

Think of it like shifting gears in a car. Low gear (Instant) is quick and easy for cruising. High gear takes more effort from the engine but handles steep hills that would stall you otherwise.

Using ChatGPT on mobile?

On iOS and Android, the model picker appears at the top of the conversation instead of in the message composer area. The reasoning options work the same way.

Turn off Higher intelligence auto-switching

ChatGPT has a setting that automatically upgrades your reasoning level when it detects a complex request. That is helpful in everyday use, but for this project you need to control reasoning manually so you can see the difference each level makes.

• Click on your profile icon in the top-right corner of ChatGPT.
• Select Settings.
• Click General.
• Find the Higher intelligence toggle.
• Note whether your toggle is currently on or off.
• Turn the Higher intelligence toggle off.

Why turn off Higher intelligence?

When Higher intelligence is turned on, ChatGPT can automatically use more reasoning for complex requests while Instant remains selected. That is great for everyday use because you get deeper thinking when you need it without manually switching.

For this project, you want to see the raw difference between each reasoning level yourself. Turning this off ensures ChatGPT only uses the exact level you choose. You will turn it back on later.

Can't find the Higher intelligence setting?

Make sure you are in Settings and then General. The toggle is under the General tab, not under Personalization or other sections. If you are on the Free or Go plan, this setting may not be available to you.

Help me find the Higher intelligence setting in ChatGPT

Check which reasoning levels your plan offers

Different ChatGPT plans give you access to different reasoning levels. Knowing what you have helps you follow the rest of this project.

• Close the Settings panel and return to your chat.
• Click the model picker in the message composer again.
• Count the reasoning levels you see on the slider.

✔️ I see Instant, Medium, and High

You are on the Plus plan. You have access to Instant, Medium, and High reasoning. This is the ideal setup for this project because you can compare all three levels side by side.

ⓧ I see more levels like Extra High or Pro

You are on a Pro, Business, or Enterprise plan. You have access to additional levels like Extra High and Pro. This project focuses on Instant, Medium, and High. You can experiment with your extra levels as a bonus after completing the project.

ⓧ I don't see a reasoning slider at all

You are on the Free or Go plan. Instead of a reasoning slider, you have a Think toggle that activates extended thinking with GPT-5.6 Luna.

You can still follow this project. Where the steps say "set the reasoning slider to Instant," use your default mode. Where they say "set it to Medium or High," activate Think mode instead. The concepts are the same, though the contrast will be less dramatic.

Why are Instant messages virtually unlimited?

Instant responses use less computing power, so ChatGPT offers them with virtually no cap. Medium and High responses cost more computing effort, so they count toward a limited weekly allowance. This is why choosing the right level matters. You do not want to burn through your High allowance on tasks that Instant handles perfectly.

You have found the reasoning slider, turned off auto-switching, and know exactly which thinking levels you have access to. Next up, you will put the slider to the test and discover why reasoning level matters more than you think.

Watch Instant Fall Short

You found the reasoning slider and took manual control of how hard ChatGPT thinks. Now it is time to put that control to the test.

In Project 2, you learned how to craft stronger prompts. But a well-crafted prompt is only half the equation. In this step, you will send a complex, detailed prompt on the fastest reasoning setting and see what comes back.

In this step, get ready to:

• Send a detailed analysis prompt to ChatGPT with the reasoning slider set to Instant.
• Read the response and identify areas that feel rushed or generic.
• Save the response as a baseline for comparison.

Start a new chat on Instant

• In your ChatGPT window, click the model picker in the message composer.
• Confirm the reasoning slider is set to Instant. This is the default and fastest setting.
• Start a new chat by clicking New chat in the sidebar.

On the Free or Go plan?

If you do not see the reasoning slider, use your default mode for this step. The concepts are the same, though the contrast between responses will be less dramatic than on Plus.

Send a complex analysis prompt

Your prompt needs to be complex enough to expose the limits of fast, lightweight thinking. You will ask ChatGPT to analyze a real decision you are facing, with multiple options, trade-offs, and a personalized recommendation.

• Think of a real decision you are currently facing. This could be a career move, a major purchase, a learning path, or a life change. Pick something you genuinely care about.
• Copy the prompt template below, replace the placeholder with your real decision, and send it in your new chat:

I need to make a decision about [[DECISION="your actual decision, e.g. whether to switch careers or which certification to pursue"]]. Analyze at least three options. For each option, evaluate the pros, cons, time investment, potential risks, and long-term impact. Then recommend which option is best for me based on what you know about me, and walk me through your reasoning step by step.

Why use a real decision?

A real decision gives you a gut check. You already have context and opinions about the topic, so you will know immediately whether ChatGPT's analysis feels useful or shallow.

The prompt also says "based on what you know about me." This pulls in your Custom Instructions and Memory from Project 1, so the recommendation should feel personal, not generic.

Read and save the response

Before you start reading, take a moment to predict. Given how detailed and specific your prompt was, how thorough do you expect the analysis to be?

• Read through ChatGPT's response from start to finish.
• Look for areas where the analysis feels thin. Are the pros and cons generic? Does it skip over risks? Does the recommendation feel rushed rather than reasoned? Is it making good use of what it knows about you from your Custom Instructions and Memory?

You probably noticed at least one area where the response felt surface-level. The prompt was strong. You followed everything from Project 2. The gap you just found is not a prompting problem. It is a thinking problem.

Why did Instant fall short?

Instant prioritizes speed. It answers quickly without extended reasoning, which works well for simple questions. But a multi-step analysis with trade-offs, risks, and personal context needs more thinking time than Instant provides.

Reasoning level is a lever your prompts alone cannot pull. No matter how good your prompt is, the quality ceiling depends on how much thinking power you aim it at.

• Save this response as your baseline. Take a screenshot, copy the text into a note, or keep this chat open. You will compare it against better responses in the next step.

Response actually seems pretty good?

If Instant gave a decent answer, your decision might not be complex enough to expose the gap. Try a decision with more competing options or higher personal stakes.

Even if the response looks fine now, save it. The contrast with higher reasoning levels in the next step will make the differences clear.

Still not sure what to look for?

You have your baseline. Next, you will send the exact same prompt with more thinking power and see how dramatically the output changes.

Dial Up the Thinking

In the last step, you sent a strong, detailed prompt on Instant and got back a shallow response. The problem wasn't your prompt.

ChatGPT skipped extended reasoning because Instant prioritizes speed over depth. Now you'll send the exact same prompt on Medium and High to see how much the output improves when you let ChatGPT think longer.

In this step, get ready to:

• Send your analysis prompt on Medium reasoning and compare it to your Instant baseline.
• Send the same prompt on High reasoning and compare all three responses.
• Identify which response you'd actually trust for your real decision.

Send your prompt on Medium reasoning

Medium reasoning gives ChatGPT more time to structure its thinking before responding. It sits between the speed of Instant and the depth of High.

• Start a new chat in ChatGPT.
• Open the model picker in the message composer and set the reasoning slider to Medium.
• Paste the exact same decision analysis prompt you used in the previous step. Do not change a single word.
• Send the prompt and watch for a thinking indicator that may appear before the response.

On the Free or Go plan?

Activate Think mode where these instructions say Medium or High. The concepts are the same, though the contrast between responses may be less dramatic.

• When the response arrives, read through it carefully.
• Compare the Medium response to your saved Instant response from the previous step.
• Notice whether the pros and cons are more specific, whether risks are explored in more detail, or whether the recommendation feels better reasoned.

Why keep the prompt identical?

If you changed even one word, you couldn't tell whether improvements came from the reasoning level or the prompt change. Keeping it identical isolates the one variable you're testing: how hard ChatGPT thinks.

Medium looks similar to Instant?

The difference between Instant and Medium can be subtle, especially for simpler decisions. The bigger jump comes when you try High in the next substep.

Help me understand why my Medium response looks similar to Instant.

Send your prompt on High reasoning

High reasoning is designed for complex, multi-step work. ChatGPT spends noticeably more time thinking before it responds.

• Start another new chat in ChatGPT.
• Open the model picker and set the reasoning slider to High.
• Paste the exact same decision analysis prompt one more time, word for word.
• Send the prompt. This response will take longer to arrive than Instant or Medium.

• When the response arrives, read through it carefully.
• Compare this High response to both your Instant and Medium versions.
• Look at how the analysis is structured, how specific the trade-offs are, and whether the recommendation addresses your personal situation.

What's the thinking indicator?

On Medium and High, you may see a progress indicator before the response appears. This shows ChatGPT is reasoning through the problem step by step before writing its answer. The longer the indicator runs, the more thinking is happening behind the scenes.

Compare all three responses side by side

Before you put all three responses together, predict which reasoning level produced the response you'd trust most to help you make this decision. Hold that prediction.

• Open all three chat conversations so you can switch between them.
• Look at which response gave the most thorough analysis of your options.
• Check which response used your Custom Instructions and Memory to tailor the recommendation to your specific situation.
• Decide which response you would actually trust to help you make this decision in real life.

The High response should show noticeably deeper analysis, more nuanced trade-offs, and more specific reasoning tied to your situation. The prompt was identical every time. The only variable was how hard ChatGPT was allowed to think.

Why does this matter?

In Project 2, you learned that better prompts produce better results. This step proves that reasoning level is a completely independent lever. Even a perfectly written prompt produces a shallow result when ChatGPT isn't given time to think.

Knowing how to write good prompts AND knowing which reasoning level to aim them at is what separates a casual ChatGPT user from someone who gets consistently excellent results.

You've proven that reasoning level changes output quality dramatically, even when the prompt stays exactly the same. Next up, you'll discover when Instant is actually the smarter choice and how to stop overthinking every question.

Learn When Speed Wins

You just proved that higher reasoning in ChatGPT produces dramatically better analysis on a complex task. The temptation is obvious: set the slider to High for everything and never look back.

Higher reasoning is slower and counts toward a limited usage allowance. Instant messages are virtually unlimited, so knowing when to save your reasoning budget is as important as knowing when to spend it.

In this step, get ready to:

• Test a quick factual question on both High and Instant reasoning to see the speed difference.
• Turn Higher intelligence auto-switching back on so ChatGPT can decide when deeper reasoning is needed.

Ask a simple question on High reasoning

Not every question benefits from extended reasoning. Some have a single correct answer that does not improve with more thinking time.

• Start a new chat in ChatGPT.
• Set the reasoning slider to High.
• Type What year was the Eiffel Tower built? and send it.
• Watch for the thinking indicator that appears before the answer.

Notice how long ChatGPT takes to answer a question with exactly one correct answer. That thinking time is the cost of High reasoning, and for a simple factual lookup, it adds no value.

Try the same question on Instant

Before you test this, predict: will ChatGPT give a different answer on Instant? Will it arrive faster or slower than the High response?

• Start a new chat.
• Set the reasoning slider to Instant.
• Send the same question: What year was the Eiffel Tower built?.

The answer is identical. But it arrived noticeably faster, with no thinking indicator and no waiting.

For a straightforward factual question, the extra reasoning on High added delay without adding value. The prompt was the same, the answer was the same, and only the speed changed.

Why save your reasoning budget?

Instant messages are virtually unlimited. Higher reasoning levels count toward a limited usage allowance.

Using Instant for simple tasks preserves your allowance for the complex questions that genuinely benefit from deeper thinking.

Turn Higher intelligence back on

Now that you understand the difference between Instant and higher reasoning, you can let ChatGPT make this choice automatically. The Higher intelligence setting handles this for you.

• Return to Settings in ChatGPT.
• Select General.
• Find the Higher intelligence toggle and turn it on.

What does Higher intelligence do?

When Higher intelligence is turned on, ChatGPT can automatically switch from Instant to a higher reasoning level when it detects a complex request.

This automatic reasoning does not count toward your manually selected reasoning usage allowance. You can still override it any time by selecting a reasoning level yourself.

You now have the best of both worlds: fast answers when speed is right, and deeper reasoning when complexity demands it.

Don't see the Higher intelligence toggle?

Make sure you are on the General tab inside Settings. The toggle may not appear on the Free or Go plan.

help me find the Higher intelligence setting in ChatGPT

You now know when to let ChatGPT think hard and when to keep it quick. Next up, you will discover the specialized tools that let ChatGPT do things that reasoning alone cannot.

Discover ChatGPT's Built-In Tools

You can now control how hard ChatGPT thinks with the reasoning slider. But thinking effort is only half the picture.

ChatGPT also has built-in tools that do things reasoning alone cannot: pulling live web data, crunching spreadsheets, and creating images. Choosing the right tool is just as important as choosing the right reasoning level.

In this step, get ready to:

• Trigger ChatGPT's Search tool with a live question and see inline citations from the web.
• Trigger Data Analysis and Image Generation to process data and create a visual.
• Build a personal AI Decision Playbook mapping your common tasks to the right reasoning level and tool.

Trigger Search with a live question

ChatGPT's Search tool pulls real-time information from the web that is not in its training data. It works like a smarter version of a web search, returning concise results with source links right inside the conversation.

• Start a new chat at chatgpt.com.
• Ask ChatGPT a question about something that happened this week. For example, try "What were the biggest news stories today?" or "What is the current price of" followed by a stock or product you follow.

Read the response carefully. You should see small numbered citations woven into the text, each linking to a real web source.

How do I know Search activated?

When ChatGPT uses Search, you will see inline citations (small numbered markers) in the response text. Clicking a citation reveals the source URL.

If the response has no citations, your question might not have required live web data. Try asking about something more time-sensitive, like today's headlines or a live stock price.

Not seeing any citations?

• Make sure your question requires information that changes over time. A question like "What is Python?" can be answered from training data and will not trigger Search.
• Try a more specific current-events question, such as asking about a sports result, a product release, or a news event from this week.

Help me trigger ChatGPT's Search tool with a live question

Trigger Data Analysis and Image Generation

Data Analysis runs code in a secure environment to process numbers and create visualizations. Image Generation creates images from text descriptions. You can trigger both from a single prompt.

Before you send the next prompt, think about this: how many distinct tool activations do you expect to see from one message?

• Start a new chat.
• Send this prompt to ChatGPT:

Create a sample CSV of monthly household expenses across six categories for the past three months. Then analyze it, show me which category had the biggest increase, and generate an image of a pie chart showing the spending breakdown.

Watch the response arrive. You should see two separate tool activations happen in sequence, not one.

What's happening behind the scenes?

First, ChatGPT's Data Analysis tool activates. You will see it running code in a secure environment to generate the CSV data and calculate which category increased the most.

Then, ChatGPT's Image Generation tool creates the pie chart. The finished image appears directly in the conversation. These are two different capabilities working one after the other, triggered by your single prompt.

• Confirm that you see a code execution step (Data Analysis) and a generated pie chart image (Image Generation) in the response.

Missing the pie chart or the code execution?

• If you see analysis text but no pie chart image, ChatGPT may have created a text-based chart instead. Try sending a follow-up: "Now generate an image of the pie chart."
• If you do not see any code execution indicator, ChatGPT may have estimated the analysis instead of computing it. Try starting a new chat and re-sending the prompt.

Help me trigger Data Analysis and Image Generation in ChatGPT

Build your AI Decision Playbook

Everything you have learned about reasoning levels and tools comes together here. You will ask ChatGPT to build a personalized reference table that maps your real tasks to the right capability.

• Start a new chat.
• Send this prompt:

Based on what you know about me from my Custom Instructions and Memory, help me build an AI Decision Playbook. List my 10 most common types of tasks and recommend the best reasoning level (Instant, Medium, or High) and the best tool (regular chat, Search, Data Analysis, Image Generation, or Deep Research) for each one. Format it as a reference table I can save and reuse.

Why does ChatGPT already know my tasks?

This prompt asks ChatGPT to use your Custom Instructions and Memory from Project 1. Because you already taught ChatGPT about yourself, it can suggest tasks that match your actual work and interests instead of generating a generic list.

• Review the playbook ChatGPT generates. Read through each row and edit any tasks that do not match your real workflow.
• Save the playbook somewhere accessible. Copy it into a note app, a document, or keep it in this ChatGPT conversation for easy reference.

Make it yours

The playbook ChatGPT generates is a starting point. Add tasks it missed, remove ones that do not apply, and adjust reasoning levels based on what you saw in the earlier steps. The more accurately this reflects your actual work, the more useful it becomes.

Before you wrap up, look at your playbook one more time. Do you think every row has the right reasoning level and tool combination, or would you change any based on what you experienced in this project?

Your playbook should have 10 rows, each mapping a real task to a specific reasoning level and a specific tool. If any row still feels generic, edit it now so it reflects something you actually do.

Playbook feels too generic?

• If the tasks do not match your real work, your Custom Instructions or Memory may need updating. Try sending a follow-up: "These tasks don't match what I actually do. Here are three things I do every week: [list them]. Rebuild the playbook around my real workflow."
• If the reasoning levels all say the same thing, ask ChatGPT to justify each choice: "For each row, explain why you chose that reasoning level over the alternatives."

Help me improve my AI Decision Playbook

Secret mission

Go Deep

You have been controlling how hard ChatGPT thinks. Now it is time to unleash its most powerful research capability. Deep Research spends minutes browsing dozens of sources and delivers a full cited report you can download and keep.

Clean Up Your Resources

Clean Up Your Resources

This project used your existing ChatGPT subscription and created no additional paid resources. Your subscription continues regardless of whether you keep or delete anything from this project.

Decide whether to keep your conversations and playbook, tidy things up, or delete everything from this project.

Resources you used:

• Conversations from this project stored in your ChatGPT sidebar (Instant, Medium, High, Search, Data Analysis, Image Generation, and Deep Research chats).
• AI Decision Playbook saved in your chosen location (a ChatGPT conversation, a note app, or a doc).
• Higher intelligence auto-switching setting, currently turned ON in Settings > General.
• One Deep Research task consumed from your monthly allowance (Secret Mission only). This cannot be recovered.

Keep everything running

No action needed. Choose this if you want to keep all your conversations, your AI Decision Playbook, and your current settings as they are.

• Your conversations from this project remain in your ChatGPT sidebar for future reference.
• Your AI Decision Playbook stays wherever you saved it.
• Higher intelligence auto-switching remains ON, so ChatGPT will automatically use more reasoning effort when it detects a complex request.

Pause - I'll come back to this later

Keep your AI Decision Playbook and clean up the practice conversations to reduce sidebar clutter. Your playbook is the main deliverable from this project.

• Keep your AI Decision Playbook conversation (or the external document where you saved it). This is your reusable reference.
• Delete the practice conversations (Instant, Medium, High comparisons and tool-trigger chats) from the ChatGPT sidebar by hovering over each conversation and clicking the delete option.
• Keep the Deep Research conversation if you want to revisit that report later.
• Leave Higher intelligence auto-switching ON. This is the recommended setting for everyday use.

Delete - I don't want to use this again

Remove all project conversations and materials. Your ChatGPT account and subscription are unaffected.

• Delete each conversation from this project by hovering over it in the ChatGPT sidebar and selecting the delete option. This includes your Instant, Medium, High, Search, Data Analysis, Image Generation, Deep Research, and AI Decision Playbook conversations.
• If you saved your AI Decision Playbook outside of ChatGPT (in a note app, doc, or other tool), delete it from that location.
• If you downloaded your Deep Research report (Markdown, Word, or PDF), delete the file from your computer.
• The reasoning slider resets per conversation, so there are no slider settings to undo. Higher intelligence auto-switching is a personal preference. Leave it ON (recommended) or turn it off in Settings > General.

That's a Wrap!

That's a Wrap!

Nice work! You just went from treating ChatGPT as a single tool to understanding it as a toolkit with different thinking speeds and specialized capabilities.

You've learned how to:

• Use the reasoning slider to control how hard ChatGPT thinks, and prove that reasoning level changes output quality dramatically even when the prompt stays exactly the same.
• Recognize when Instant is the right choice to save time and your reasoning budget, and when Medium or High is worth the wait.
• Trigger ChatGPT's built-in tools deliberately (Search, Data Analysis, and Image Generation) and build a personal AI Decision Playbook that maps your most common tasks to the right reasoning level and tool combination.
• Secret Mission: Ran a real Deep Research task, compared its output to a regular High-reasoning chat, and experienced ChatGPT's most powerful research capability firsthand.

Ready to quiz yourself?