Tokens and costs: what does a long conversation use?

1. 1

   Why can a subscription still ask you to wait?

   Imagine a usage-limit notice appears after several revisions. That does not, by itself, mean you have incurred an extra charge. Read whether it is a temporary limit or an offer to buy access. Price and permitted usage are separate conditions.

   A subscription provides access under a plan. An API lets software send requests and often bills by usage. Even with one provider, the app subscription and API billing may be separate.

   App limits can depend on the model, messages, attachments, features and other conditions. Do not assume they are simply a token allowance. Check your own plan and reset notice. The arithmetic below covers only text input and output in an API.

2. 2

   Characters and tokens are different units

   Tokens are pieces used to process text. A word may be one piece or several, and spacing and punctuation matter too. There is no universal one-character-to-one-token conversion.

   

   The tokenizer determines the split. Counts and differences between languages depend on the text and model. A result from one Korean sentence is not a rule for every Korean document.

   Use a short, non-sensitive sample and the intended provider's counting tool. Another provider's count is only an estimate. Recorded API usage can also include instructions and conversation history, so keep estimates separate from measured usage.

1. 3

   Calculate input and output separately

   Input is what you send; output is what the model returns. Rates are often different. A short answer to a long report can still use substantial input, while a short question can produce a costly long answer.

   For practice only, assume $2 per million input tokens and $10 per million output tokens. These are invented exercise rates, not a provider quote. With 2,000 input tokens and 300 output tokens, calculate each side separately.

   Do not omit the million-token denominator. The $0.007 total applies to this single hypothetical request. It excludes tax, currency conversion, tools, caching and discounts; check those separately for a real estimate.

   One request at invented exercise rates

   Item	Calculation	Cost
   Input	2,000 ÷ 1,000,000 × $2	$0.004
   Output	300 ÷ 1,000,000 × $10	$0.003
   Total	$0.004 + $0.003	$0.007

2. 4

   A short follow-up can carry a long history

   A two-word follow-up may still include earlier messages and answers if the application sends that history with it. Developers can inspect the actual request and usage records.

   If the second request contains 3,000 input tokens including history and returns 200 output tokens, our exercise rates give $0.006 plus $0.002: $0.008. The two requests together cost $0.015.

   Applications manage history differently, and caches may change API billing. This example teaches why request count and context size both matter; it is not a billing formula for every chat app.

3. 5

   Fitting in context does not guarantee a good answer

   The context window is the information a model can work with in one request. A larger desk holds more pages, but does not guarantee every line gets equal attention. Capacity is different from quality.

   

   Select material relevant to the task, but preserve important exceptions and conditions. Keep titles, dates and page references so you can trace the answer back to its source.

   If a conversation loses track of earlier constraints, make a checked summary of established facts and open questions before moving to a new conversation. Do not assume the new one knows the old context.

This is a learning check, not a test you must pass. Use the explanations to understand the ideas.

1. Q1At the exercise rates, keep 2,000 input tokens and increase output to 600. What is the cost?

   Write your thoughts (optional)

   Write your thoughts (optional)

   Your writing is not sent anywhere. It disappears when you leave or reload this page, so copy anything you want to keep.

   Read the answer and explanation

   $0.004 input plus $0.006 output equals $0.010. It is not twice the original $0.007 because input did not grow. These are invented exercise rates.

   I read and understood the explanation

   This is your own review, not an automatic score.

2. Q2An app asks you to try again later. Can you calculate an extra charge from your message count?

   Write your thoughts (optional)

   Write your thoughts (optional)

   Your writing is not sent anywhere. It disappears when you leave or reload this page, so copy anything you want to keep.

   Read the answer and explanation

   No. The notice alone does not establish any extra charge. Check the app's usage and billing terms rather than applying our hypothetical API rates.

   I read and understood the explanation

   This is your own review, not an automatic score.

3. Q3After a long document, you ask for a one-line summary but see high input usage. What would you check?

   Write your thoughts (optional)

   Write your thoughts (optional)

   Your writing is not sent anywhere. It disappears when you leave or reload this page, so copy anything you want to keep.

   Read the answer and explanation

   Check whether the document and earlier messages were sent again. The latest user message is not necessarily the whole input. Inspect measured usage before concluding the service is wrong.

   I read and understood the explanation

   This is your own review, not an automatic score.

1. 6

   Your turn: a budget without making a purchase

   Assume 20 requests per day for 30 days, all the same size as the first example. The daily cost is $0.14 and the monthly total $4.20. That is exercise arithmetic, not a forecast of your bill. Record the rate units and assumed request count beside the total so a later revision can distinguish volume changes from price changes.

   Double only the output and calculate again. Input stays fixed, so the total does not double. This helps you see the effect of requesting only the detail you need without removing explanations necessary to understand the result.

   Before real API use, record the model, rates and effective date, then inspect usage for a small request. Distinguish a budget alert from an enforced spending stop: an alert may not stop calls. API implementation belongs in the developer course.