MEASURED DATA
AI COSTS MORE
IN YOUR LANGUAGE
The same meaning, priced in 25 languages. Greek pays 2.32× what English pays to say exactly the same thing — and almost nobody paying that bill knows it.
LanguageGPT-4GPT-4o
- GreekΕλληνικά5.42×2.32×
- HungarianMagyar2.40×2.02×
- UkrainianУкраїнська3.16×1.91×
- CzechČeština2.42×1.86×
- PolishPolski2.12×1.78×
- RomanianRomână1.97×1.68×
- Hindiहिन्दी5.35×1.67×
- TurkishTürkçe2.17×1.64×
- Japanese日本語2.28×1.63×
- FinnishSuomi2.07×1.62×
- Hebrewעברית3.89×1.58×
- VietnameseTiếng Việt2.62×1.54×
- Korean한국어2.43×1.50×
- ItalianItaliano1.70×1.49×
- RussianРусский2.55×1.49×
- Arabicالعربية3.18×1.45×
- GermanDeutsch1.70×1.38×
- IndonesianBahasa Indonesia1.66×1.36×
- SwedishSvenska1.66×1.35×
- PortuguesePortuguês1.55×1.26×
- SpanishEspañol1.45×1.25×
- FrenchFrançais1.51×1.24×
- DutchNederlands1.63×1.19×
- Chinese中文1.64×1.16×
- EnglishEnglish1.00×1.00×
Method: the real BPE tokenizers — cl100k_base for the GPT-4 generation and o200k_base for GPT-4o and later — are run over a parallel corpus of five texts per language. The token counts are exact, not estimated. The translations are ours and are kept in full in the repository, because a wordier rendering would cost more tokens and you should be able to check ours.