What is a token?
A token is the unit an LLM reads and generates. It's usually a chunk of a word, not a whole word. The word "calculator" might be one token, while "tokenization" could split into two or three. As a rule of thumb, 1 token ≈ 4 characters ≈ 0.75 words for English.
This counter uses a character-and-word heuristic that lands within a few percent of the real count for typical English. For exact billing, each provider ships its own tokenizer, and Claude and GPT don't tokenize identically.
Why token count matters
Every API call is billed by tokens, split between input (what you send) and output (what the model returns). Knowing your token count up front lets you predict cost, stay under context limits, and decide when to trim a prompt or switch to a cheaper model.
Run these tokens through one API
Claude, GPT-5 and 50+ models behind a single OpenAI-compatible endpoint. Pay per token, $1 free to start.
Start FreeRelated tools
AI API Cost Calculator — project your monthly bill across models and request volume.
FAQ
How accurate is this counter?
Within a few percent for English prose. Code and non-Latin scripts tokenize differently, so treat it as an estimate, not an exact bill.
Do Claude and GPT count tokens the same?
No, they use different tokenizers. The same text can yield slightly different counts on each.
How many tokens per word?
About 1.3 tokens per English word on average.