In AI pricing, a token is a small chunk of text that a language model reads or writes, such as a short word, part of a word or a punctuation mark. One token is roughly four characters of English, or three-quarters of a word. Model vendors charge per million tokens, with separate prices for input (what you send) and output (what the model writes).
Key takeaways
- One token ≈ four characters ≈ three-quarters of an English word. 1,000 tokens ≈ 750 words.
- Bills have input, output, reasoning and cached tokens. Output usually costs about five times input at the big three vendors.
- Models have no memory between calls, so conversation history is resent and paid for again.
- Tokens are your cost unit. Most businesses shouldn't make customers think in them.
Let me run the numbers on my own book.
The Subscription Playbook is about 50,000 words across its ten chapters, or about 313,000 characters. To an AI model, that's somewhere around 75,000 tokens.
Feed the whole book to Claude Sonnet 5.5 and you'd pay about 15 cents. Ask it to write the whole book back to you and you'd pay about 75 cents. (Writing it the first time took me considerably longer, and cost a lot more coffee.)
That little sum is the heart of AI pricing. So let's start with the word itself.
What does token mean in AI?
A token is a chunk of text that a language model reads or writes. It might be a whole short word ("cat"), part of a longer word ("subscr" + "iption"), a number or a punctuation mark.
OpenAI's rule of thumb is that one token is about four characters of English, or roughly three-quarters of a word.
| Text | Roughly |
|---|---|
| 1 token | 4 characters |
| 100 tokens | 75 words |
| 1,000 tokens | 750 words, about a page and a half |
| 1 million tokens (1 MTok) | 750,000 words |
Other languages and code often use more tokens per word, so treat these as estimates.
The four kinds of tokens on an AI bill
- Input tokens: everything you send the model. Instructions, the customer's question, documents and the conversation so far.
- Output tokens: everything the model writes back. These cost more.
- Reasoning tokens: the "thinking" a reasoning model does before it answers, usually billed as output.
- Cached input tokens: repeated input the vendor has stored, billed at a steep discount.
Here's the part most people miss. A model has no memory between calls. Ten messages into a chat, the eleventh call sends all ten previous messages again, and you pay for them again.
How are tokens priced?
Model vendors quote a price per million tokens, with separate rates for input and output. As at 27 October 2026, Claude Sonnet 5.5 and GPT-6.1 Sol both list at $2 per million input tokens and $10 per million output tokens. Small models such as GPT-6 Luna and Claude Haiku 5.5 cost $0.10 and $0.50.
Two things stand out. Output costs about five times more than input at OpenAI, Anthropic and Google. And the gap between a small and a frontier model is about 100 times.
To turn tokens into dollars, see cost per token, or compare every model in the LLM API pricing comparison.
AI Token Cost CalculatorEstimate monthly LLM API spend across models.Open the free tool →Token vs credit: what's the difference?
A token is the model vendor's unit of cost. An AI credit is the app's unit of price. Most AI apps buy tokens from a model vendor and sell credits, plans or outcomes to their own customers.
How tokens connect to the book
In Chapter Four I wrote about usage-based pricing and its "feast or famine" risk, using my old telecommunications job as the example. Tokens are the same idea as phone minutes. And just like phone companies, the smartest AI businesses don't make customers think in the raw unit. They sell per document, per conversation or per minute, and keep tokens as their internal cost. More on that choice in usage-based pricing.
Estimate your token bill
Enter your prompts and volumes, and the free AI Token Cost Calculator estimates your monthly spend across models.
What to do next
Count the tokens in your three most common tasks, including instructions and history. Then price them with the AI Token Cost Calculator.
For the full picture, including caching, batch discounts and whether to charge customers per token, read token-based pricing explained. If you're deciding how to price the product on top, start with the AI pricing guide.
Back to the subscription and pricing glossary.
Frequently asked questions
Is an AI token the same as a crypto token?
No. In AI pricing, a token is a unit of text processed by a language model. A crypto token is a digital asset on a blockchain, and a payment token is a stand-in for card details used by payment processors. They share a name and nothing else. If you're pricing an AI product, you only need the first meaning.
Do images and audio use tokens too?
Often, yes. Many models convert images, audio and video into tokens before processing them, and each vendor publishes its own conversion rates. Others price these inputs per image, per minute or per character instead. ElevenLabs, for example, sells speech in credits tied to characters, and Deepgram prices transcription per audio minute.
What is MTok?
MTok is shorthand for one million tokens, the standard unit for model prices. When a vendor lists $2 per MTok input, it means $2 for every million tokens you send. One million tokens is roughly 750,000 words, or about fifteen books the length of The Subscription Playbook.
Why do reasoning models use more tokens?
Because they 'think' before answering, writing out intermediate steps that you usually don't see. Those reasoning tokens are typically billed as output, the expensive side of the bill. A short question to a reasoning model can therefore cost several times more than the same question to a standard model, even at the same listed price.
Get both books free
The Subscription Playbook (ebook + audiobook) and Feed a Starving Crowd, instantly. In Australia? Claim the free hardcover.
Robert Coorey, MBA
Robert is the author of The Subscription Playbook and Feed a Starving Crowd and a Director at BPAG Accountants in Sydney. More about Robert.
Sources
- OpenAI Help: What are tokens and how to count them
- OpenAI: API pricing (checked 27 Oct 2026)
- Anthropic: Claude API pricing (checked 27 Oct 2026)
Facts and prices checked 27 October 2026. Spotted something out of date? Tell us.