Does my text leave the browser?

No. Tokenization runs entirely in your browser; the page makes no network request with your text.

Why do models count the same text differently?

Each model family uses its own tokenizer vocabulary. The same sentence can differ by 20 percent or more between families, which changes what a prompt costs.

How accurate are the estimated counts?

Estimates use per-family character ratios and are typically within about 10 percent. Counts marked with a tilde are estimates; unmarked counts are exact.

Free tool

How many tokens is that?

Paste text, pick a model, get the count. Exact OpenAI-family tokenizers in the browser, calibrated estimates for the rest.

Last reviewed July 17, 2026

ModelText

Tokens (GPT 5.6 Sol): ~0; calibrated estimate
Words: 0
Characters: 0

How this is computed

OpenAI-family models tokenize exactly with their real encodings, loaded in your browser on demand; nothing you paste leaves the page. Other families use per-family characters-per-token ratios calibrated against provider documentation, marked with a tilde. The converter tab applies the same ratios to word counts.

Frequently asked

More free tools

What will your workload cost?Will it fit the context window?

Built from these features

Multi-Model Chat

GPT, Claude, Gemini, Grok, Llama: pick mid-conversation, cheapest-first.

Ready to get to work?

Free to start. No credit card. Never trained on your data.

Start free