EN
Sign inGet started
Glossary

Token

What is Token?

A token is the smallest unit of text a language model processes, usually part of a word, a short word or a punctuation mark. Models read inputs and generate answers token by token, and the size of the context window is also measured in tokens. Because providers charge for input and output tokens, tokens directly determine costs and limits: long instructions and extensive background knowledge increase consumption.

How KI.ordo helps

KI.ordo shows token consumption and costs per model and lets you set a monthly token limit.

Learn more →: Token

Topics

  • LLM
  • Costs
  • Basics

Try KI.ordo

Agents, flows and templates in one workspace – free 24-hour demo.

Start demo
← All terms

Related terms

LLM Costs

The costs of large language models are usually billed per token, i.e. text units roughly the size of a syllable to a word; input and output tokens are priced separately, with output generally more expensive. Larger models cost more per token than small, fast models such as Claude Haiku, which are often sufficient for simple tasks. Prompt caching lowers costs when the same long context is sent repeatedly. Concise instructions and carefully selected knowledge help as well.

  • Costs
  • Tokens
  • +1

Large language model (LLM)

A large language model (LLM) is a neural network trained on very large amounts of text, enabling it to understand and generate language. It works by calculating the most likely next token, step by step. LLMs write, summarise, translate, answer questions and generate code; well-known examples include Claude, GPT, Gemini and Llama. Without access to current sources, they only know what was in their training data.

  • Language models
  • Generative AI
  • +1

Cache-augmented generation (CAG)

Cache-augmented generation (CAG) is an approach in which a language model does not search for knowledge with every question. Instead, a manageable knowledge base is loaded into its context window in advance and that state is cached, so answers need no separate retrieval step. The term was popularised by a late-2024 research paper as an alternative to retrieval-augmented generation (RAG). CAG suits small, stable knowledge bases; for large or frequently changing data, RAG usually remains more practical.

  • RAG
  • Context window
  • +1