Paste a prompt. We estimate tokens for four model families using empirical chars-per-token ratios (plus adjustments for code, unicode and repeated whitespace), then project cost across premium, standard and budget pricing tiers — input, output and blended. Pure client-side heuristic; nothing leaves your browser.
4 chars/token baseline (cl100k)
Heuristic only — within ~5% of a real BPE tokeniser on English prose, ~10% on code. Use it for budgeting and context-window checks; not for billing.