You’re probably thinking of Belgium and Switzerland where 97 would be - as far as I understand - “nonante-sept”
See this other StackExchange on the topic (and geographical exceptions to the octante/huitante (!)/nonante usage): https://french.stackexchange.com/questions/187/quelles-parti...
80 - vuitanta
Luckily it is only numbers 50 to 99 that work that way. 31 is simply "enogtredive", as in "one and three tens". A hundred is "hundred", not "fems", as it could be (five twenties).
70: soixante-dix (FR) "sixty-ten". Septante (BE) is literally seventy (seven decades), which is much better.
And so on and so forth for 80 (octante in BE) and 90 (nonante).
Edit: fixed seventy
Nope. That’s switzerland, belgians completely illogically have kept the 20-based naming here.
So do Belgians do it 100% like the french? Or so they mix a bit of FR and CH?
Octante as far as I know has been dead for a while and is not used anywhere, in Switzerland or elsewhere.
Let's prove that English is insane by making a literal translation of that Wikipedia definition in French:
> Littéral traduction, direct traduction ou mot-pour-mot traduction, est le traduction de un texte fait par traduire chaque mot séparément, sans regarder à comment le mots sont utilisé ensemble dans un phrase ou phrase.
A similar literal translation from english to english of 97 would be nine-tens-seven. English has some irregularities in the teens but it's decimal all the way through. French switches from decimal to vigesimal.
I'm trying to show that using a method that can only produce nonsense out of a given source is not the best way to demonstrate/prove _anything_ about the source. At best, it proves that literal translation is garbage, which is something everyone knows (except at tech giants, apparently).
There is only one way to translate "quatre-vingt-dix-sept" in English while preserving its meaning: "ninety-seven".
Pointing out the decimal <-> vigesimal switch (and the historical debates around it, for example) or the fact that the two systems cohabit would be far more useful and interesting than constructing a strawman in order to conclude that a language is "insane".
If you perform the same literal translation to 57, it doesn't look "insane". When you say the method "can only produce nonsense", I think you're just flat-out wrong.