“On the FrontierCode "Extended" tier, Claude Opus 4.8 scored 51.8%, GPT-5.5 scored 44.8%, and Claude Opus 4.7 scored 43.2%.”