Nous Research's open-source Nomos 1 AI model scored 87/120 on the notoriously difficult Putnam math competition, ranking ...