Free Token Counter for GPT, Claude, Gemini & Llama
Paste your text, see the token count instantly. No upload, no signup.
LLM APIs bill by tokens, not characters or words, and every model splits text into tokens differently. This tool counts tokens locally in your browser using the same byte-pair-encoding family OpenAI's models use, then estimates counts for Claude, Gemini, and Llama from that base count.
Nothing you paste is uploaded or stored: the count runs entirely on your device.
| Model | Estimated tokens | Context window | Fits? |
|---|---|---|---|
| GPT (OpenAI) | — | 1.05M | — |
| Claude (Anthropic) | — | 1M | — |
| Gemini (Google) | — | 1.05M | — |
| Llama (Meta) | — | 1M | — |
The GPT count is exact (tiktoken-compatible BPE, the same encoding OpenAI's own tokenizer uses). Claude, Gemini, and Llama figures are estimates scaled from the GPT count, since those vendors don't publish a tokenizer that runs in a browser; expect them to be within roughly 5-15% of the real count. Context windows shown are each vendor's current flagship model, verified against their own docs.
Tokens per word, roughly
If you don't have text to paste yet, here's a rough planning reference for English prose. Code, non-English text, and unusual formatting all shift this: code typically runs 10-20% more tokens per character than prose, since GPT-family tokenizers were built around natural-language vocabulary.
| Text | ≈ tokens (English) |
|---|---|
| 1 word | ~1.3 tokens |
| 1 sentence (~15 words) | ~20 tokens |
| 1 paragraph (~100 words) | ~130 tokens |
| 1,000 words | ~1,300-1,500 tokens |
| 1 page (~500 words) | ~650-750 tokens |
These are averages for standard English prose (Cohere's tokenizer guide and OpenAI community benchmarks both put it around 1.3 tokens per word). Paste your actual text above for an exact count instead of estimating from this table.
Estimating an API bill, not just a fit?
This tool counts tokens; it doesn't price them; per-token API rates change often enough that we track them separately.
See real per-clip AI generation pricing across vendors →Common questions
What is a token?
A token is the chunk of text a language model actually processes, usually a few characters or part of a word rather than a whole word. In English, one token is roughly 4 characters or about three-quarters of a word.
How many tokens is 1,000 words?
About 1,300 to 1,500 tokens for standard English prose, since one word averages roughly 1.3 tokens. Simple, common words can run close to 1 token each; longer or less common words split into 3 or more.
Why does Claude count tokens differently than GPT?
Every model family trains its own tokenizer on its own vocabulary, so the same sentence splits into a different number of pieces per model. GPT's tokenizer (o200k_base) is public and exact to run in a browser; Anthropic hasn't published one that runs client-side, which is why Claude's number here is an estimate rather than an exact count.
Why do token counts matter?
Most AI APIs charge per token and cap how many tokens fit in a single request or context window. Counting tokens before you send a prompt tells you whether it fits, and roughly what it will cost.
Is this counter exact for every model?
It's exact for GPT models, which use a public tokenizer. Claude, Gemini, and Llama don't publish a tokenizer that runs in a browser, so those numbers are estimates based on the GPT count.
Is my text sent anywhere?
No. Tokenization happens entirely in your browser using a local library; the text you paste never leaves your device.