New benchmark study results show leading AI models, including ChatGPT, Claude, and Gemini, still lag humans in visual math ...