Domain specific models outperform large vision language models on cytomorphology tasks

Large vision-language models (LVLMs) show impressive capabilities in image understanding across domains. However, their suitability for high-risk medical diagnostics remains unclear. We systematically evaluate four state-of-the-art LVLMs and three domain-specific models on key cytomorphological benchmarks: peripheral blood cell classification, morphology assessment, bone marrow cell classification, and cervical smear malignancy detection. Performance is assessed under zero-shot, few-shot, and fine-tuned conditions. LVLMs underperform significantly: the best LVLM achieves a zero-shot F1 score of 0.057 {+/-} 0.008 for malignancy detection--near random (0.039)--and only 0.15 {+/-} 0.01 in few-shot. In contrast, domain-specific models reach up to 0.83 in accuracy. Even after fine-tuning, a dedicated hematology model outperforms GPT-4o. While LVLMs offer explainability via text, we find the visual-language grounding unreliable, and the morphological features mention by the model often do not match the single cell properties. Our findings suggest that LVLMs require substantial improvements before use in high-stakes diagnostic settings. Key findingsO_LILVLMs perform poorly on cytomorphology tasks, often near chance level and far below domain-specific models. C_LIO_LIEven after fine-tuning, LVLMs lag behind domain-specific models. C_LIO_LIWhile LVLMs provide textual justifications, these often reflect generic descriptions rather than image-specific morphological features. C_LI