Token Counter
Exact token counts for OpenAI models, plus calibrated per-model estimates for Claude, Llama, Gemini and Mistral.
Processed locally in your browser. Nothing you paste is sent anywhere.
- Uses the actual tokenizer for OpenAI models, so the count matches what the API will bill and budget.
- For Claude, Llama, Gemini and Mistral, uses a characters-per-token ratio calibrated per family — not one generic ±10–20% guess applied to every non-OpenAI model.
- Token, word and character counts side by side, updating as you type.
- Runs in your browser — prompts and documents aren't uploaded.
Why this is private
- Processed locally in your browser. Nothing you paste is sent anywhere.
- Open your browser's DevTools → Network tab while using this tool — you won't see a request carrying your input, because it never leaves your device.
- Nothing you enter is uploaded to a server or shared with us — the processing happens entirely on your device.
Overview
Counts tokens for a piece of text — an exact count using OpenAI's real tokenizer for GPT models, or a calibrated per-model-family estimate for Claude, Llama, Gemini and Mistral, none of which publish a fully open tokenizer.
How to use it
- 1Paste your text.
- 2Choose the tokenizer matching your target model — the dropdown groups exact tokenizers separately from approximate ones.
- 3The token, word and character counts update live.
Tips
- Approximate tokenizers are most accurate for plain English prose — expect more drift for code, non-English text, or heavy punctuation/whitespace, where character-based estimates diverge more from any real BPE tokenizer.
Frequently asked questions
What people use it for
Check a prompt fits the context windowConfirm a system prompt plus context is well under the model's limit before sending.
Estimate the cost of a requestTake the token count to the AI API Cost Calculator for a full monthly cost estimate, instead of multiplying by hand.
Trim a prompt to a budgetEdit and watch the count until it's under a target.
Size chunks for embeddingsMake sure each chunk stays under the embedding model's input limit.
Compare a prompt's rough length across model familiesSwitch the tokenizer to see how the same text is estimated for Claude vs. GPT vs. Gemini.
Related tools
All ai & llm toolsContext Window CalculatorSee how much of a model's context window your text uses.AI API Cost CalculatorEstimate monthly LLM API costs from token volume.JSONL ValidatorValidate a JSONL (JSON Lines) file line by line.JSONL FormatterFormat or minify a JSONL file, with malformed lines isolated instead of failing the whole file.
Ready to use Token Counter?
Token Counter