The field of medical intelligence is rapidly advancing with a focus on multimodal understanding, integrating text, images, and other modalities to improve clinical decision-making and patient outcomes. Recent developments emphasize the need for comprehensive evaluation frameworks to assess the performance of large language models and vision-language models in real-world medical tasks. Researchers are introducing new benchmarks and datasets, such as multimodal question answering datasets for Traditional Chinese Medicine and STEM disciplines, to evaluate the capabilities of these models in specific domains. Another area of focus is the development of clinically trustworthy models for medical image diagnosis, with benchmarks like DrVD-Bench assessing the clinical visual reasoning capabilities of vision-language models. Moreover, there is a push towards creating more efficient and deployable models for tasks like automated radiology report generation and placenta analysis, leveraging techniques such as contrastive distillation and knowledge transfer from large foundation models. Noteworthy papers in this area include CSVQA, which introduces a diagnostic multimodal benchmark for evaluating scientific reasoning in STEM disciplines, and ReXVQA, the largest benchmark for visual question answering in chest radiology, demonstrating AI performance that exceeds expert human evaluation in certain tasks. These advancements highlight the potential for artificial intelligence to support and augment clinical expertise, paving the way for more accurate, efficient, and reliable medical practices.