Automatic Number Plate Recognition (ANPR) is a component of intelligent transportation systems. ANPR systems rely on Optical Character Recognition (OCR) algorithms to extract text from license plate images. However, their accuracy can be hindered by environmental factors and the diversity of license plates. This study investigated the use of multimodal large language models with vision capabilities to overcome these limitations. Multimodal large language model performances in license plate character recognition tasks were evaluated by comparing them to a single-modal OCR algorithm. These multimodal models achieved higher accuracies and lower character recognition error rates, suggesting that their contextual understanding coupled with their natural language processing improves the accuracies of their OCR tasks. Therefore, the integration of large language models within OCR pipelines may improve the accuracy of OCR models.