Does the best AI model's lead come from quantization, or the model?

Models and Hub profiles the arms call

Arms

Judged scores

Scenario × arm  (click a cell)

Recent  (click a row)