Tokens
The unit of input and output for large language models - roughly 0.75 of a word in English. LLM providers price by tokens; metering AI apps usually does too.
Last updated: 2026-05-10
Definition
Tokens are not characters and not words - they are subword pieces produced by a tokenizer. "Hello, world!" is about 3 tokens; "I am building an AI product" is about 7. Different models tokenize differently. For OpenAI, use tiktoken to count locally; for Anthropic, use their tokenizer endpoint. Vevee supports tokens as a first-class unit on limit groups.
Example
Charging a user 0.5 cents per 1k input tokens + 1.5 cents per 1k output tokens, marked up from your provider cost.
Related terms
Usage-based pricing
A pricing model where the customer's bill scales with their usage of the product, instead of a flat subscription. Sometimes called "consumption pricing" or "pay-as-you-go".
GlossaryMetered billing
Charging customers based on what they actually consumed in a billing period - tokens, image renders, agent runs - instead of a flat subscription fee.
GlossaryComposite event
A single tracked event (one vevee.track call) that, via match rules, increments more than one limit group at the same time. The default model in Vevee.