README / README.md
mkbadeniyi's picture
Update README.md
8dd30a0 verified
|
Raw History Blame Contribute Delete
1.57 kB
metadata
title: README
emoji: ✂️
colorFrom: gray
colorTo: red
sdk: static
pinned: false
short_description: Africa's open AI models, smarter, faster and lighter

African languages shouldn't pay a token tax. So we cut it. Scissors cut an Igbo sentence from 25 tokens to 13.

Quantile Labs

Africa's open AI models, made smarter, faster and lighter.

AI reads text in small pieces called tokens. Words in African languages get split into more of them than English words, so the same answer is slower and costs more to run. We call this the token tax, and we cut it.

We start from open models built in Africa and teach them to read their own languages in bigger pieces. We choose our quality tests before training, publish the results, and release open weights that run offline on everyday hardware. One country after another, one language after another.

Models

Model Built on Languages Links
WazoCore N-ATLaS, Nigeria Yoruba, Hausa, Igbo Model · Code · Page
Next One country after another

quantilelabs.com · Our bet · All models · GitHub