On this page
Definition
A token is the unit of text an LLM processes — typically 3-4 characters or roughly three-quarters of an English word. Prompts, replies and pricing are all measured in tokens.
Simple explanation
Instead of reading whole words, an LLM reads pieces called tokens. 'ChatGPT' might be split into 'chat' and 'GPT'. Numbers, punctuation and rare words often use multiple tokens.
You never need to count tokens by hand — but knowing they exist explains why some prompts are surprisingly expensive.
Why it matters
Everything about running an LLM — cost, speed, context limits — is measured in tokens. If you build with AI, tokens are the meter.
How it works
- 1TokeniseText is split by a tokenizer into vocabulary IDs.
- 2FeedThe model processes IDs, not characters.
- 3GenerateIt outputs one token at a time.
- 4BillYou are charged per input token and per output token.
Real examples
Products named for illustration only. Inclusion is not an endorsement.
- OpenAI pricingCharged per million input and output tokens, priced separately.
- Anthropic pricingSame input/output split, often with cache discounts.
Advantages
- Consistent unit across models and providers.
- Enables usage-based pricing at fine granularity.
Limitations
- Token counts are hard to estimate by eye.
- Different tokenizers count the same text differently.
- Costs balloon on long documents if not planned for.
Common misunderstandings
- ClaimOne token equals one word.RealityOften it does, but many words split into multiple tokens.
- ClaimOnly inputs cost money.RealityOutputs usually cost more per token than inputs.
Frequently asked questions
How many tokens is a page of text?
Roughly 500-800 tokens per standard page.
Why do outputs cost more?
Generating a token is more compute-intensive than reading one.
Can I reduce token usage?
Yes — concise prompts, targeted retrieval and short reply formats all help.
The Tool Money Lab perspective
The teams that treat tokens as a budget line item are the same teams that ship AI features on time and on cost. Ignoring token accounting is the fastest way to build a product that cannot scale.
Conclusion
Tokens are the atoms of language models: the unit of reading, writing, cost and speed.
Design for token efficiency and the economics of AI features stay predictable.