d1-ultra tops the Artificial Analysis ranking because the model more often says 'I don't know'

The model that more often stays silent took first place: d1-ultra scored 65 points on the Artificial Analysis Decision Index versus 57 for Jev 1.14. The share of honest 'I don't know' answers on AA-Omniscience-Decision questions rose from 2% for d1 to 85%, while accuracy fell from 41% to 32%.