Token (in the context of LLM)
A token is the minimum unit of text in language models. What is it, why is Russian text more expensive than English and how it affects the work with AI tools.
A token is the minimum unit of text that a language model works with. This is not always a word: one Russian word is often broken into 2-4 tokens, and the Russian text on average consumes 1.5-2 times more tokens than English of the same length. The model reads, processes and generates text with tokens.
For professionals, tokens are important at two points:
In working with the API, the cost of the request is considered in tokens, so a long prompt with excess water is literally more expensive.
In a content strategy, dense and specific text without water is more efficiently processed by the model when quoting in AI responses.
Related terms
Fine-tuning (model training)
Fine-tuning – additional LLM training on highly specialized data. What is it, what is different from RAG and what approach is more important for GEO promotion.
Artificial intelligencellm
LLM (Large Language Model) is a neural network behind ChatGPT, Claude and Gemini. How it works, why it’s wrong and what it means for SEO and content marketing.
Artificial intelligenceRAG (Retrieval-Augmented Generation)
RAG is an AI architecture in which a model searches for data from external sources before responding. Learn how it works and why it’s important for GEO.
Artificial intelligenceLLM hallucination
LLM hallucination is when an AI generates compelling but false text. What is it, why does it arise, and how does it affect SEO and AI search?
Did not find a term?
Need an answer for your website?
Describe the task and we will suggest what to check first.