Nemotron-3.5-Embed-8B-BF16

A person can use it to build semantic search and retrieval-augmented question-answering systems over text and document images.

Availability

Hugging Face release dated 10/20/2026

Pricing model

Not established

Deployment

Model examples use a local checkpoint and NVIDIA GPU-accelerated systems

Privacy

Not established

Licence

OpenMDW License Agreement, version 1.1 (OpenMDW-1.1)

• The model generates embeddings from text and image inputs for multimodal retrieval. ^[1]

  The model generates dense vector embeddings from text or image inputs, enabling multimodal retrieval, semantic search, and (agentic) RAG workflows.

• Text retrieval was evaluated across 34 languages and image retrieval across 18 languages. ^[1]

  For text retrieval, the model was evaluated across 34 languages: English, Arabic, Assamese, Bengali, Bulgarian, Chinese, Danish, Dutch, Finnish, French, German, Hindi, Hinglish, Indonesian, Italian, Japanese, Korean, Malay, Marathi, Nepali, Norwegian, Persian, Portuguese, Romanian, Russian, Spanish, Swahili, Swedish, Tamil, Telugu, Thai, Ukrainian, Urdu, and Vietnamese.

• The model outputs a 4096-dimensional embedding vector and supports truncated embedding sizes. ^[1]

  Typically, the embedding model is used first to embed (vectorize) the whole corpus (document images or text chunks), and embeddings are stored in a vector database associated to its raw content (image or text).

• The source gives a Hugging Face release date of 10/20/2026. ^[1]
• The source provides Transformers, Sentence Transformers, and vLLM usage examples. ^[1]

  Create a Python 3.12 environment for the Transformers and Sentence Transformers examples:

• The source says the model is ready for commercial or non-commercial use under the OpenMDW-1.1 download terms. ^[1]

• Create expressive speech from a short voice sample Daily brief 2026-10-27