Welcome the Nvidia Llama Nemotron Nano VLM to Hugging Face Hub (huggingface.co)

NVIDIA has introduced the Llama Nemotron Nano VL, an advanced 8B Vision Language Model (VLM) now available on the Hugging Face hub. This state-of-the-art model is designed for intelligent document processing, showcasing high accuracy in extracting and understanding complex information from documents such as invoices, contracts, and diagrams. With exceptional performance on the OCRBench v2 benchmark, Llama Nemotron Nano VL outperforms its competitors, making it particularly valuable for industries like finance, healthcare, and legal sectors that require efficient document automation. The model employs a Vision Transformer (C-RADIOv2-VLM-H) for powerful visual processing and multimodal capabilities, enabling it to accurately handle a variety of document layouts, including text, tables, and charts. Leveraging a robust training data strategy that includes over 1.5 million samples, Llama Nemotron Nano VL excels in critical tasks such as text recognition, table extraction, and visual reasoning. Additionally, its compatibility with NVIDIA NeMo allows users to further fine-tune the model with custom datasets. This combination of high accuracy, low model footprint, and efficient deployment makes it a strong contender for enterprises seeking to streamline document workflows and enhance business intelligence capabilities.