Word ↔ token estimator
A quick two-way converter between words and tokens. Handy for back-of-the-envelope budgeting from a word count (e.g. "a 500-word answer"). It uses a sensible average ratio — real counts vary by model and content, so calibrate with your own text below.
Calibrate with your own text (real tokenizer)
Words
—
Tokens (real)
—
Real tokens/word
—
Default ratio: ~1.33 tokens per word (≈ 0.75 words/token, ≈ 4 characters/token) for typical English. Code, other languages, numbers, and rare words run higher. The calibration box uses the real o200k_base tokenizer (exact for OpenAI; a close proxy for others).
More free token tools
TokensToken counterPaste text, pick a model, and see exact tokens, characters, and chars/token — every token as a colored chip with its ID.CostLLM API cost calculatorReal token counts → dollar cost, split into input vs output, for the model you pick. Paste text to auto-count input.CostModel cost comparisonOne prompt + expected output length, costed side by side across every model in a sortable table.TokensTokenizer visualizerSee exactly how text splits into tokens, color-coded, with hover to inspect each token's ID and boundaries.TokensContext window checkerPaste a prompt, pick a model, and find out whether it fits the context window — and how many tokens are left.ExplainerFrom prompt to outputA clickable pipeline explaining how an LLM turns your prompt into output — tokenization, embeddings, transformer, sampling, and back.