ModelsOct 31, 20261 min readOpen-Weight AI Trails Gemini by 44.8 Points on HLE Reasoning - FourWeekMBA0SaveShareContentsWhy it mattersTopicsEcosystem AccessOpen WeightsSelf-hostable & Fine-tunableCategoryModelsIndustry DomainThe Core DevelopmentOpen-Weight AI Trails Gemini by 44.8 Points on HLE Reasoning FourWeekMBAWhy it mattersOpen-Weight AI Trails Gemini by 44.8 Points on HLE Reasoning FourWeekMBA Reported by Google News — Google DeepMind on October 31, 2026.Topic dossiers Google DeepMind & Gemini