LLM Token Counter
Paste any text to see exactly how many tokens it uses under OpenAI tokenizers (o200k_base and cl100k_base), plus character, word and sentence counts. Estimates for Claude and Gemini are included too. Counting happens locally in your browser — nothing is sent anywhere.
🔒 100% private — files never leave your browser
Exact OpenAI BPE ranks — the same tokenization the API uses.
- Tokens
- 0
- Words
- 0
- Characters
- 0
- Sentences
- 0
Also: 0 paragraphs · 0 characters without spaces · ballpark estimate 0 tokens (4 chars/token).
Estimated input cost
Compare API costs →GPT-5 input
$0.00
Claude Sonnet input
$0.00
Gemini Flash input
$0.00
Based on one request with the selected token count and current built-in rates.
How to use the Token Counter
- 1 Paste or type your text — prompt, document or code.
- 2 Pick a tokenizer: o200k_base (GPT-4o / GPT-5) or cl100k_base (GPT-4 / GPT-3.5), or an estimate mode for Claude and Gemini.
- 3 Read the exact token count, characters, words and sentences — updated live as you type.
Frequently asked questions
What exactly is a token?
Models read text as tokens — common English words are one token, rarer words split into pieces. As a rule of thumb, 1 token ≈ 4 characters ≈ 0.75 words of English. Token counts determine both context limits and API pricing.
Are the counts exact?
For OpenAI tokenizers they are exact — the same BPE ranks the API uses, running locally via Tiktoken. Claude and Gemini do not publish tokenizers, so those modes use o200k_base as a calibrated estimate (typically within ±10–20%).
Why do tokens matter?
Context windows, per-request pricing, rate limits and embedding sizes are all measured in tokens. Checking a prompt before sending avoids truncated contexts and surprise bills.