LLM Token Counter

A token is the chunk of text a model reads and bills by. It can be a word, part of a word, a space plus a word, or punctuation. Here's one sentence split by o200k_base, OpenAI's current published tokenizer:

Notice the leading spaces: “ ship” and “ship” are different tokens. Common English words are one token each, while contractions and punctuation split off. You pay for input tokens on every call, and the context window caps how many fit.

Every model family ships its own tokenizer with its own vocabulary, so the same text becomes a different number of tokens. Some providers publish their tokenizer, some expose a counting API, and some do neither. This tool is upfront about which is which:

Published by the providers themselves, for English text:

• OpenAI: about 4 characters or three-quarters of a word per token, so about 100 tokens per 75 words. Source
• Google Gemini: about 4 characters per token; 100 tokens is about 60 to 80 English words. Source
• DeepSeek: about 0.3 tokens per English character and 0.6 per Chinese character. Source

All three say the same thing in different words: other languages, code and unusual formatting cost more tokens per character. When the number matters, count it rather than estimate it.

• Claude token counter →
• GPT token counter →
• Gemini token counter →
• Llama token counter →
• DeepSeek token counter →
• Mistral token counter →