Fine-Tuning Dataset Formatter
Turn question and answer pairs into a JSONL training file for OpenAI, Llama or Gemini.
Token Counter
Estimate how many tokens your text costs before you send it.
0Est. tokens
0Words
0Characters
0Chars / token
0Tokens / word
Context window
How much of a window this text fills, before any answer is generated.
What this would cost
Enter your model's rates per million tokens. Both are editable.
This is an estimate, not a tokenizer. A real count comes from the model's own vocabulary file, which differs between providers and model generations. Expect to be within about ten percent on ordinary English prose, and wider than that on code, non-Latin scripts and heavy punctuation. Confirm with your provider's token counting endpoint before you rely on a hard limit.
Quick Tool Overview
The Token Counter estimates how many tokens your text costs before you send it, alongside the word and character counts and what that volume would cost at rates you can edit. A token is roughly three quarters of an English word — common words are one token, rarer ones split into pieces.
- What it does
- Estimates token count and cost for a block of text
- You give it
- Your prompt, document or code
- You get back
- An estimated token count, word and character counts, and a cost
- Best for
- Sizing context windows and sanity-checking an API bill
About the Token Counter
Models charge by the token, not the word, and a token is roughly three quarters of an English word — common words are one token, rarer ones split into pieces, and punctuation, code symbols and emoji each take their own. Paste your text and this gives you an estimated token count, the word and character counts beside it, and what that volume would cost at rates you can edit.
Be clear about what this is: an estimate, not a tokenizer. A real count comes from the model's own byte-pair vocabulary, which is megabytes of data and differs between providers and model generations. This uses a heuristic calibrated on English prose and code, which lands within about ten percent for normal text. That is accurate enough to size a context window or sanity-check a bill, and not accurate enough to bet a hard limit on — if you need the exact number, count it through the provider's own token counting endpoint.
Key features of the Token Counter
Estimate within about ten percent
A heuristic calibrated on English prose and code, accurate enough to size a context window.
Words and characters alongside
The three counts sit together, which makes the roughly-four-characters-per-token relationship easy to sanity-check.
Editable cost rates
Type the current per-million rates from your provider and see what the text would cost to send.
Honest about accuracy
A real count comes from the model's own byte-pair vocabulary, so the tool tells you to verify through the provider's token counting endpoint before betting a hard limit on it.
Nothing is transmitted
Prompts and documents are measured in your browser, so unreleased content stays local.
How to use the Token Counter
-
1
Paste your text
Drop in a prompt, a document or a block of code. The estimate updates as you type.
-
2
Read the estimate
You get estimated tokens, words, characters and the ratio between them, plus how much of a context window it fills.
-
3
Check the cost
Set the input and output rates for the model you use and see what the text would cost per call.
-
4
Verify before you commit
For a hard limit, confirm with your provider token counting endpoint. This is a fast estimate, not the tokenizer itself.
Frequently asked questions
Within roughly ten percent for ordinary English prose. Code, heavy punctuation, non-Latin scripts and long unusual words all tokenize less predictably, so treat those figures as a wider range.
A real tokenizer needs the model vocabulary file, which is megabytes and different for each provider and model generation. Downloading that on every page load would break the promise that this site stays fast and fully client-side.
About 0.75 words for English prose, so 1,000 tokens is roughly 750 words. Code runs denser, closer to 0.5 words per token, because symbols and indentation each cost a token.
No. Output is typically several times the price of input, which is why long generated answers cost far more than long prompts. Both rates are editable here so you can price them separately.
Reviews of the Token Counter
What people who use the Token Counter think of it. Reviews come from visitors and are posted as written.
No ratings yet
Yours would be the first.
- 5 0 reviews rated 5 stars
- 4 0 reviews rated 4 stars
- 3 0 reviews rated 3 stars
- 2 0 reviews rated 2 stars
- 1 0 reviews rated 1 star
0 reviews
Newest first
No reviews yet
Be the first to rate the Token Counter. It takes half a minute and no account.
Write the first reviewMore AI tools
Prompt Template Builder
Fill a prompt template from spreadsheet rows and get one prompt per row.
GPU VRAM Calculator for LLMs
Work out how much GPU memory a local LLM needs to run or fine-tune.
AI Crawler Blocker
Generate robots.txt rules to block GPTBot, ClaudeBot, CCBot and other AI crawlers.
Prompt Injection Checker
Scan untrusted text for prompt injection patterns and hidden characters before a model reads it.
JSONL Validator
Validate a JSONL file line by line, and check fine-tuning datasets before you upload.
Function Schema Generator
Write valid tool-calling JSON schemas without hand-editing braces.
Few-Shot Example Builder
Turn input and output pairs into a clean, consistent few-shot prompt block.