Nearby lessons
19 of 30Tokens & Cost
Concept in One Line
In real API usage, you are charged based on the number of tokens processed, not by questions, time, or answers.
Why This Concept Is Important
- Helps students understand real-world AI cost.
- Very important for developers and startups.
- Common interview question for AI roles.
Key Points to Remember
- API cost depends on token count.
- Both input tokens and output tokens are charged.
- Longer prompts mean higher cost.
- Longer answers mean higher cost.
- Token cost is model-dependent.
What Is Charged in API Usage?
When you use GPT API, you pay for:
- Input tokens (your prompt)
- Output tokens (model response)
Total cost = input tokens + output tokens
Simple Example
Example 1: Short Prompt
Prompt: Explain AI
Few input tokens
Short output
Low cost
Example 2: Long Prompt
Prompt: Explain AI with history, examples, advantages, disadvantages, interview questions
Many input tokens
Long output
Higher cost
More words = more tokens = more money
Why Companies Care About Tokens
- Millions of users means millions of tokens.
- Small increase per request can create a huge cost.
- Token optimization saves real money.
Simple Flow
Prompt Length -> Token Count -> API Computation -> Cost
Interview / Exam Points
Q1: How is GPT API pricing calculated?
GPT API pricing is calculated based on the number of input and output tokens processed.
Q2: Is API cost based on number of questions?
No. It is based only on token usage.
Common Confusions to Avoid
- Cost is not per question.
- Cost is not per minute.
- Cost is not per response.
- Cost is per token.
One Line to Remember
In APIs, every token has a cost.
Quick Self-Check
For Students
API cost depends on token usage - Yes / No?
Short prompts always cost more - Yes / No?
(Correct answers: Yes, No)