NVIDIA NeMo Retriever, a collection of world-class microservices, sets the industry standard for multimodal RAG, by achieving 1st place across all visual document retrieval leaderboards–ViDoRe V1, ViDoRe V2, and MTEB VisualDocumentRetrieval.
NVIDIA’s new Llama NeMo Retriever ColEmbed model is fine-tuned for query-to-document retrieval—think text queries matched to images and is designed for multimodal RAG systems that handle text, charts, tables and infographics.
Get started:
📺 Watch our video to learn more about how NeMo Retriever leads the way in visual document retrieval.
🔎Research–Try the NeMo Retriever ColEmbed model for free on Hugging Face.
🖥️Enterprises–for production-ready, commercial models, visit Explore Retrieval Models | Try NVIDIA NIM APIs.
🛠️Developers looking to jump-start the AI workflows with customizable reference applications can get started with our RAG blueprint.