A new benchmark has evaluated the visual perception abilities of advanced artificial intelligence models. The results show that none of the models have achieved 60% accuracy in recognizing visual information. This suggests that AI systems still struggle with interpreting and understanding visual data, often making errors during the initial image-reading stage. The findings highlight the need for further research and development in this area. This matters because the ability of AI to accurately perceive and understand visual information is crucial for applications in areas such as autonomous vehicles and medical imaging.