Embedding
Nemotron 3.5 Embed
Nemotron 3.5 EmbedNemotron 3.5 Embed makes enterprise text and visual content searchable, helping agents retrieve relevant information from documents, tables, charts, and diagrams.
Model details
Nemotron 3.5 Embed is an open multimodal embedding model for enterprise retrieval. It creates unified representations of text, images, and mixed text-image content, helping applications find relevant information across documents, tables, charts, diagrams, and page images even when answers are expressed visually rather than in text.
Available in 1B and 8B sizes, Nemotron 3.5 Embed offers options to balance serving efficiency and retrieval quality. Use it to power multilingual search and agentic RAG, and adapt it for domain-specific data and retrieval needs.
