Token
In one sentence
The basic unit LLMs use to process text. Both model context length and API call costs are calculated in tokens.
A token can be a single character, part of a word, or punctuation, depending on the model's tokenizer. As a rough estimate, one Chinese character typically corresponds to 1 to 2 tokens, though exact counts depend on each model's pricing documentation.
API costs are usually priced separately for input tokens and output tokens, with outputs generally costing more than inputs. When budgeting AI costs, businesses must calculate based on the actual tokens consumed by each task.
Example
Having an AI read a 5,000-word report and generate a 1,000-word summary requires input and output tokens to be billed separately.
Related terms
- Context window
- The maximum total amount of text an LLM can process at once, including prompts, reference materials, chat history, and its own generated responses.
- Large Language Model (LLM)
- AI models trained on massive volumes of text that can understand and generate text. Doubao, DeepSeek, ChatGPT, and Claude are all powered by large language models.
- API
- Interfaces that allow software programs to interact with each other. Enterprise systems and AI agents use APIs to access capabilities such as large language models, search, image generation, and publishing.
Further reading
AI News
Chinese LLMs Capture 46% of US Enterprise Token Traffic
Driven by price gaps up to 35x, Chinese models now handle 46% of US enterprise API tokens on OpenRouter, pushing teams toward dynamic multi-model routing.
Jul 14, 2026 · 3 min read
AI News
NVIDIA Vera Rubin: Building the Token Factory, Not Just a Server
NVIDIA's Vera Rubin shifts enterprise AI to industrial scale, cutting token costs by 90% and turning compute into a utility for autonomous agents.
Mar 17, 2026 · 3 min read
