The field of biological vision and visual understanding is rapidly evolving, with a focus on developing models that can learn and represent complex biological concepts. Recent research has highlighted the importance of large-scale training data and hierarchical contrastive learning in achieving emergent properties in biological vision models. These models have been shown to exhibit remarkable accuracy in various biological visual tasks, such as habitat classification and trait prediction, and can capture long-range spatial dependencies across entire rows or columns in images. However, studies have also revealed limitations in current models, including biases in vision language models and a lack of hierarchical knowledge in large language models. To address these challenges, new frameworks and benchmarks have been proposed, such as strip-aware spatial perception for fine-grained bird image classification and dynamic benchmarks for species discovery using frontier models. Notable papers include BioCLIP 2, which achieved extraordinary accuracy in biological visual tasks, and TerraIncognita, a dynamic benchmark for evaluating state-of-the-art multimodal models for species discovery.