A side-by-side look at where each model gets things right — and where they still stumble.
See model scores →
Each card shows one isiZulu sentence and the number of words every model got wrong.