Locked learning resources

Join us and get access to thousands of tutorials and a community of expert Pythonistas.

Unlock This Lesson

Locked learning resources

This lesson is for members only. Join us and get access to thousands of tutorials and a community of expert Pythonistas.

Unlock This Lesson

Usage, Limits, and Tokens

Sarah Jane Ross

AI for Beginners: Use Agents With Confidence Sarah Jane Ross 02:50

Every AI app has limits on how much you can use it, and some of your everyday choices use up those limits much faster than others. Once you know what makes one request heavier than another, you can get the same work done and run into limits less often.

How Usage Is Counted

Models read and write text in tokens, which are small chunks of text, often a piece of a word. Pick a sample to see how it splits into tokens:

Everyday words are often one token each, while web addresses, numbers, and other languages break into more, smaller pieces.

Locked learning resources

Join us and get access to thousands of tutorials and a community of expert Pythonistas.

Unlock This Lesson

Already a member? Sign-In

Locked learning resources

The full lesson is for members only. Join us and get access to thousands of tutorials and a community of expert Pythonistas.

Unlock This Lesson

Already a member? Sign-In

00:00 Usage, Limits, and Tokens. Every AI app has limits on how much you can use it. Some of your everyday choices use up those limits much faster than others.

00:11 Once you know what makes one request heavier than another, you can get the same work done, and run into limits less often.

00:18 Models read and write text in tokens. Tokens are small chunks of text, often a piece of a word. Here’s a short request, split into tokens. Everyday words are often one token each.

00:32 But web addresses and numbers break into more, smaller pieces. So does text in other languages.

00:40 Everything the model reads counts, and that’s more than the messages you type. It includes the app’s own instructions. It includes the conversation so far.

00:48 And it includes any file you uploaded earlier in the chat. In a long conversation, that adds up quickly. So the same one-line question can be light or heavy.

00:59 In a fresh chat, the model reads very little. At the end of a long chat, with a big file uploaded earlier, it reads all of that, too. The question is the same, but the second one uses much more.

01:11 How usage turns into limits depends on how you pay. With a subscription, you pay a fixed amount. You get a usage limit that resets after a while, often every few hours or every week.

01:22 Some apps count the number of messages. Others count how much the model reads and writes. With those, a long conversation or a big file uses up the limit faster than a quick question.

01:34 When you pay per use, you pay for the tokens themselves, or for credits that stand in for them. App builders often work this way. So does an API key, which is a pass that lets an app use a model on your account.

01:46 Either way, the same choices make the difference. A heavier request uses more of your limit, or it costs more. Tokens the model writes use more than tokens it reads.

01:57 That includes the thinking it does before it answers. Many apps offer this as a thinking or reasoning mode.

02:04 With more thinking, the model works through the problem step by step before it replies. Every step is more writing. So the answer can stay the same, while the writing behind it grows.

02:15 Thinking helps with hard problems. Building a project schedule with lots of constraints is one. Checking a tricky calculation is another. But for a quick question, thinking uses more of your limit without making the answer better.

02:29 So the heaviest requests are the ones where the model reads a lot, writes a lot, or both. Long conversations and subagents add up too, and the next two lessons cover them.

02:39 Below this video, you can pick a sample and see how it splits into tokens. You can also move a thinking dial, and watch the writing behind the same answer grow.

Become a Member to join the conversation.