What is a token in an LLM?

The basic unit of text a model processes — typically a word or fragment of a word; models read and generate text token by token, and tokens are the unit for pricing and context limits.

Why do tokens matter for cost?

AI usage is almost always priced per token, so longer inputs and outputs cost more — verbose prompts repeated at scale can drive large bills.

How many words is a token?

Roughly three-quarters of a word in English on average, though it varies — tokens are word-pieces, so a long or unusual word may be several tokens.

Growth Marketing Glossary

Token (LLM)

to·kennoun

The unit AI reads, writes, and bills in - roughly a word-piece. Tokens determine cost, speed, and how much fits in the context window.

Schematic — text split into tokens

Term: Token (LLM)
Is: A chunk of text — a word or word-piece
Models process: Text token by token
Tokens measure: Cost, limits, context window size

Forms & parts of speech

token · noun

A unit of text for an AI model.

"Long prompts cost more because pricing is per token - and they eat into the context window."

Definition in plain terms

A token is the basic unit of text that a large language model works with. Rather than reading whole words or characters, models break text into tokens, which are typically words or fragments of words - a common rough estimate is that a token is about three-quarters of a word in English.

The model reads its input as a sequence of tokens and generates its output one token at a time. Tokens matter practically for three reasons. First, AI usage is almost always priced per token, so longer inputs and outputs cost more.

Second, the context window - how much the model can consider at once - is measured in tokens. Third, generation speed relates to how many tokens are produced. Understanding tokens demystifies AI cost, limits, and performance.

Why it matters to growth leaders

Tokens are the currency of working with AI, so understanding them is essential for any growth team using AI tools at scale.

Because cost is per token, the length of prompts and outputs directly drives the expense of an AI-powered feature, content pipeline, or support system - a verbose prompt repeated across millions of requests adds up.

Tokens also explain the context-window limits that constrain how much information you can feed a model at once.

For a growth leader managing AI in production, thinking in tokens helps control costs (by trimming unnecessary prompt length), set realistic limits, and forecast the economics of scaling an AI feature.

It connects the abstract experience of using AI to the concrete unit that determines what it costs and what it can handle - the kind of literacy that separates experimenting with AI from running it efficiently at scale.

Worked example. A growth team scales an AI-powered content and support feature from a pilot to production and is shocked when the AI bill balloons - and tokens explain both the cost and how to control it.

Because AI usage is priced per token, every word of the lengthy, repeated prompts and every word of the generated outputs was being billed, multiplied across millions of requests.

The verbose system prompt the team had pasted into every call - much of it unnecessary - was quietly the biggest cost driver.

The growth leader, now thinking in tokens, trims the prompts to the essential instructions, removes redundant context, and caps output length where appropriate, cutting the token count per request dramatically and the bill with it.

The same token lens clarifies the context-window limits the team had been hitting and lets the leader forecast the economics of scaling further.

Understanding tokens as the currency of AI, the growth leader turns an alarming, opaque cost into a managed, predictable one - the literacy that separates experimenting with AI from running it efficiently at scale, where every token of prompt and output has a price.

Failure modes to watch. Ignoring that AI cost is per token, so verbose prompts repeated at scale drive big bills; forgetting that the context window is measured in tokens; not trimming unnecessary prompt length in production; and failing to forecast token-based economics when scaling an AI feature.

Synonyms & antonyms

Synonyms

tokenLLM tokenAI token

Antonyms

whole-word processingcharacter

Origin & history

A token is the word-piece unit by which language models read and generate text; because usage is priced per token and context windows are sized in tokens, tokens are the practical currency of AI cost, limits, and speed.

Etymology: source.

Usage trends

Search interest for this term over the last five years:

View interest-over-time on Google Trends →

Common questions

What is a token in an LLM?: The basic unit of text a model processes — typically a word or fragment of a word; models read and generate text token by token, and tokens are the unit for pricing and context limits.
Why do tokens matter for cost?: AI usage is almost always priced per token, so longer inputs and outputs cost more — verbose prompts repeated at scale can drive large bills.
How many words is a token?: Roughly three-quarters of a word in English on average, though it varies — tokens are word-pieces, so a long or unusual word may be several tokens.

Related tools & calculators

toolRGM benchmarks compendium

Resources & people to follow

referenceWikipedia — large language model
referenceAI and growth-marketing practice
referenceRGM analysis — tokens are the currency of AI; trim prompt length and cap outputs to control cost at scale

Curated, non-competitor resources verified per term.

Related training

moduleMarketing analytics

Disciplines

Areas of marketing where token (llm) is a core concern:

AI searchGrowth strategy

Sources

trendsGoogle Trends — "llm token"