Language models read text as tokens — chunks that are roughly three-quarters of a word. Your prompt’s token count drives both the model’s context-window limit and what you pay per API call, so counting first helps you stay within limits and estimate cost.

For GPT models, yes — we use OpenAI’s public tokenizers (o200k_base for GPT-4o and 4.1, cl100k_base for GPT-4 and 3.5), so the counts match the API for the text you enter. Anthropic does not publish Claude’s tokenizer, so the Claude figure is a clearly-labelled estimate (about one token per 3.6 characters), not an exact count. Note that the count is for your text alone: a full chat API request adds a few tokens of overhead per message for role markers and formatting.

No. All counting happens entirely in your browser, so your prompt never leaves your device — it is safe to use with private or proprietary content.

Different model families use different tokenizers, so identical text can produce slightly different token counts. Switch the model to compare how each one sees your text.