New Benchmark Shows AI Models Lack Intuitive Visual Reasoning

Elorian and Scale AI released Humanity’s Sixth Sense (HSS), a benchmark with 522 questions across 288 images and 234 short video clips, designed to measure how a model interprets implied visual information. The human baseline scored an aggregate pass@1 of 93.1, while the best AI system, GPT 6 Astra, achieved 53.6. Chinese open weight models performed poorly, with GLM-5.3-flash scoring 26.8 and Kimi K3 scoring 25.5. The report also notes that the AI race is moving into space, with a space economy boom expected in the 2030s.