There’s a small dropdown sitting near your chatbot’s message box that most people never touch. In ChatGPT it’s at the top of the chat. In Gemini it’s tucked next to the prompt box. The labels say things like Auto, Fast and Thinking, and if you’re like most people, you’ve glanced at it, decided it looked like something for developers, and gone back to typing.
It’s not for developers. It’s a choice between a quick answer and a careful one, and it quietly affects almost everything the bot gives you. Leave it on the wrong setting and you’ll wait twenty seconds for a reply to a simple question — or get a shallow, half-thought-through answer to something that actually mattered.
It’s a speed-versus-thinking trade-off, not a quality ladder
Here’s the simple version of what’s going on. When you pick Fast, the chatbot answers the way a friend answers a casual question — quickly, from the top of its head, without stopping to work anything through. When you pick Thinking (or Pro, Opus, depending on the app), it pauses and reasons through the problem step by step before it writes the reply. That takes longer — sometimes noticeably longer — but the answer is usually more careful.
Auto sits between the two. It lets the app decide how much effort your question deserves. A simple “rephrase this sentence” gets the quick treatment; a tangled spreadsheet question gets the slow one.
The key thing to understand is that this isn’t a straight ladder from bad to good. The careful option isn’t better in every way — it’s slower, and on most paid plans the heavier options come with usage limits while the lighter ones don’t. Using the deepest reasoning mode for every trivial question is like hiring a structural engineer to hang a picture frame. You’ll get an answer, but you paid for it in waiting time you didn’t need to spend.
Where the menu actually hides in each app
Every app puts this control in a slightly different place, but it’s always near where you type:
ChatGPT: the model picker at the top of the conversation. Paid users can choose between GPT-5’s Auto, Fast and Thinking modes, and GPT-4o is also listed there — OpenAI brought it back after users asked for it. Free users get a default option without the same range of choices.
Gemini: a model dropdown near the prompt box on the web version, with Fast, Thinking and Pro options. On the web you can also type @ in the prompt box to bring up the model menu quickly.
Claude: a dropdown at the top of the conversation showing the current model. Haiku is the quick one, Sonnet is the everyday middle ground, and Opus is the heaviest reasoning option — which ones you see depends on your plan, with free accounts getting the lighter end of the range.
One honest caveat: the names and exact options change more often than you’d like. Companies rename models, shuffle plans and roll features out gradually. Don’t memorize the menu — understand the pattern. Nearly every chatbot now offers some version of quick, balanced and deep thinking, and that three-way split is the part worth learning.
A simple rule for which one to pick
You don’t need a flowchart. This covers nearly everything:
Leave it on Auto (or the default) for ordinary stuff. Casual questions, drafting an email, brainstorming names, explaining something you don’t understand — the app is perfectly capable of picking the right effort level here. This is where most of your chats should live.
Switch to Fast when speed matters and the stakes are low. Rewriting a sentence, getting a quick summary of something short, bouncing ideas around, checking a fact you mostly already know. If the answer comes back wrong, you’ll spot it instantly and just ask again — nothing was lost.
Switch to Thinking when accuracy matters more than speed. This is the one people underuse. Math and logic problems. Code you actually plan to run. A contract or a long document you need analyzed carefully. A decision with real consequences — money, health, legal stuff. A plan for something complicated. In these cases the extra seconds of waiting are the whole point: you’re buying deliberation, not decoration.
A useful test: ask yourself whether you’d want a smart friend to answer off the top of their head, or to sit quietly for a minute and think. If it’s the second, pick the thinking option. And if the question is about a long document, give the chatbot the file to read first, then choose the careful mode — it can only reason well about material it can actually see.
Two things the menu can’t fix
First: a deeper-thinking model is still a chatbot, and it can still be confidently wrong. Slower reasoning reduces mistakes; it doesn’t remove them. If the answer matters, check it against the source — we wrote a full guide on how to do that — regardless of which mode you used.
Second: don’t expect a model switch to rescue a chat that’s going stale. If you’ve been talking in one thread for weeks and the answers are getting vague or repetitive, that’s usually the conversation running out of working memory, not the wrong model setting. A fresh chat on the default setting will beat a tired chat on the smartest one.
And one small habit that pairs well with all of this: if you keep typing the same preferences into every chat — tone, format, context about you — stop retyping them and set them as standing instructions instead. The right mode plus standing instructions covers about ninety percent of what people fiddle with in settings.
Try it deliberately, just once. The next time you ask something genuinely hard, switch to the thinking option before you hit send. The next time you’re just chatting, use the fast one. Auto will handle everything in between — which, it turns out, is most of life.
