Token Counter
Paste anything — see tokens, words, and what that text costs on any model. 100% in your browser, nothing uploaded. Prices updated 2026-08-02.
- As input
- $0
- As output
- $0
- × 1,000 requests
- $0
- % of 128K context
- 0%
Estimate calibrated for English (~4 chars/token). Code and non-English text may run 10–60% higher.
Quick token math worth memorizing
- 1 token ≈ 4 characters ≈ ¾ of an English word. 1,000 tokens ≈ 750 words.
- A single page of text (~500 words) ≈ 670 tokens.
- A 10-page document ≈ 6,700 tokens — pennies as input on most models.
- Code, JSON, and non-English text run 10–60% more tokens than the same length of English prose.
- Output usually costs 3–5× input — a long answer costs more than a long question.
Frequently asked questions
› How accurate is this token count?
It is an estimate based on character and word statistics calibrated against common tokenizers (roughly 4 characters or 0.75 words per token for English). Real tokenizers differ slightly per model family, and code or non-English text tokenizes less efficiently. Expect the estimate to land within ±10% for English prose — close enough for cost planning, not for hard context-limit decisions.
› Is my text uploaded anywhere?
No. The counting happens entirely in your browser with JavaScript — nothing you paste ever leaves your device. That makes it safe for confidential prompts, contracts, or source code.
› Why do code and non-English text use more tokens?
Tokenizers are trained mostly on English text, so common English words compress to a single token, while code symbols, whitespace patterns, and non-Latin scripts split into more pieces. Arabic or Japanese text can use 1.5–2.5× more tokens per character than English, which directly raises API costs.
› What is the difference between input and output cost for the same text?
Providers charge separately for tokens you send (input) and tokens the model generates (output), with output typically 3–5× more expensive. This page shows both numbers for your text: what it costs to send it, and what it would cost if a model generated it.