Artificial intelligence has made impressive strides, tackling complex tasks in various fields. But how does it fare in the intricate world of electrical engineering? Researchers have developed a new benchmark called EEE-Bench, designed to test the abilities of large multimodal models (LMMs) – AIs that can process both text and images – to solve practical electrical engineering problems. EEE-Bench covers ten core electrical engineering subjects, from circuit design to electromagnetics, using real-world examples with diagrams and equations. The results are revealing: current LMMs struggle significantly. Even the best-performing model only achieved about 46% accuracy, highlighting a substantial gap between AI capabilities and human expertise in this field. Interestingly, closed-source models (like those from Google and OpenAI) generally outperformed open-source alternatives, suggesting that access to vast computational resources plays a significant role. However, even these powerful models faltered when faced with complex diagrams, showing a weakness in visual reasoning. The research also unearthed a curious phenomenon dubbed “laziness.” When given extra text information, even if misleading, the AI often ignored the accompanying images. This suggests that current LMMs might lean too heavily on text, sometimes neglecting crucial visual clues. EEE-Bench provides a valuable tool for researchers to identify the limitations of existing AI and pave the way for future advancements. While AI may not be ready to replace electrical engineers just yet, this research highlights both the challenges and the immense potential for AI to assist in solving complex real-world engineering problems.