Token
The unit an LLM reads and writes. Roughly 3-4 characters of English, or about 0.75 words. Every API bill is counted in tokens.
What it is
A model does not see letters or words. Text is first split into tokens - short chunks from a fixed vocabulary the model was trained on. Common words are one token; rare words split into several. Punctuation, spaces and code symbols are tokens too.
A rough rule for English: 1 token ~= 4 characters ~= 0.75 words, so 1,000 tokens is about 750 words. Other languages, code and numbers tokenise less efficiently.
Why it matters
Tokens are the meter. APIs charge per million tokens, split between input tokens (your prompt) and output tokens (the reply). Model limits - the context window - are stated in tokens. Throughput is measured in tokens per second.
Cost & infrastructure impact
Because output tokens usually cost 2-5x more than input tokens, the shape of your workload matters as much as its size. A chatbot with long context and short answers is input-heavy; an agent that writes code is output-heavy. The blended token price tries to fold both into one number.
Example
A support bot handling 100,000 chats/month, each ~700 input + 350 output tokens, uses ~70M input and ~35M output tokens/month. At $0.15 / $0.60 per 1M that is about $32/month; at $3 / $15 it is about $735.
Related concepts
Use it in Obolith
FAQ
How many tokens is a page of text?
A dense A4 page is roughly 500-700 words, so about 700-950 tokens.
Do all providers tokenise the same way?
No. Each model family has its own tokeniser, so the same text can be a different token count on OpenAI vs Anthropic vs Google. Differences are usually small but real.