Every text model works with smaller pieces of text called tokens. Understanding them makes it easier to compare models, manage long conversations and read API pricing without guessing.

A token is a unit of text

For text models, a token is one of the units the model processes. Depending on the language and tokenizer, a token might represent part of a word, a whole short word, punctuation, or a space joined to a word. One token is not necessarily one word.

“Make a short travel video.”This is an illustrative sentence — not an exact tokenizer breakdown.

A convenient rule of thumb for English is roughly four characters per token on average, but real counts vary. Other languages can have quite different ratios.

Input, output, and context

Input tokens cover the information sent to a model: your instruction, previous messages and any included material. Output tokens are generated in its reply. The context window limits how much material the model can work with at once; the exact accounting depends on the model and API.

YOUR PROMPT→MODEL + CONTEXT→RESPONSE

For example, a very long transcript plus a short question can consume far more input tokens than the response itself.

A simple billing example

Imagine a fictional API that charges $2 per million input tokens and $8 per million output tokens. A request with 10,000 input tokens and 2,000 output tokens costs $0.036 in this hypothetical example: (10,000 × $2 + 2,000 × $8) / 1,000,000. These are demonstration rates, not current prices for any provider.

Tokens versus credits

Tokens usually measure text-model usage. Image and video apps may instead charge credits per image, generation, second of footage, or tool operation. A subscription credit is a provider-specific unit, not a universal token. Never compare “100 credits” on two different platforms as if they were equivalent.

How to check a real token count

  1. Open a token counting tool compatible with the model you use.
  2. Paste the exact text rather than estimating from word count.
  3. For APIs, consult the provider’s usage and token-counting documentation.

Official reference: OpenAI — Understanding and counting tokens.

Primary sources

Official vendor documentation was used to verify the workflow. Some options depend on plan, region, product version or account access.