Nemotron 3.5 Embed 1B Overview

Description:

Nemotron-3.5-Embed-1B is a versatile embedding model trained by NVIDIA and optimized for multimodal and multilingual question-answering retrieval. The model can embed document pages in the form of image, text, or combined image–text inputs. Documents can be retrieved given a user query in text form. The model supports page images containing text, tables, charts, and infographics. 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. For image retrieval, it was evaluated across 18 languages: Arabic, Bengali, Chinese, English, Persian, Finnish, French, German, Hindi, Indonesian, Japanese, Korean, Russian, Spanish, Swahili, Telugu, Thai, and Yoruba. The model may support additional languages beyond these evaluation sets.

This model is ready for commercial or non-commercial use.

Nemotron 3.5 Embed 1B is released in BF16 and NVFP4 variants. This combined Model Card covers both variants. NVIDIA API Catalog provides one hosted Preview API experience using the BF16 variant.

Nemotron-3.5-Embed-1B-NVFP4 is the quantized version of the Nemotron-3.5-Embed-1B-BF16 model, developed for multimodal and multilingual question-answering retrieval. The model can embed document pages in the form of image, text, or combined image–text inputs. Documents can be retrieved given a user query in text form. The model supports page images containing text, tables, charts, and infographics. The model is quantized with NVIDIA Model Optimizer, using nvidia-modelopt v0.47.0.

License/Terms of Use:

Governing Hosted-Service Terms: Use of this trial service is governed by the NVIDIA API Trial Terms of Service
Source Model License: Use of the model is governed by the OpenMDW License Agreement, version 1.1 (OpenMDW-1.1).
Additional Information: The Ministral-3-3B-Instruct-2512-BF16 and Ministral-3-8B-Instruct-2512-BF16 base models are governed by the Apache License, Version 2.0.

Deployment Geography:

Global

Use Case:

The Nemotron-3.5-Embed-1B-BF16 is suitable for users who want to build a multimodal question-and-answer application over a large corpus, leveraging the latest dense retrieval technology. The input of the model is a text or a document image, and the output is a fixed-size embedding vector. The embedding model is a bi-encoder that supports context in textual format (e.g. the query or the OCR text of a page or a section of a document) or the image of a document page. 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). Then at inference time, the embedding model is used to embed the query. The model generates dense vector embeddings from text or image inputs, enabling multimodal retrieval, semantic search, and (agentic) RAG workflows.

Generally Nemotron-3.5-Embed-1B-BF16 and Nemotron-3.5-Embed-1B-NVFP4 models share the embedding space and can be used interchangeably, but we recommend validating retrieval quality on a representative sample before switching the models.

Release Date:

Build.NVIDIA.com: 10/20/2026 via model page
Hugging Face (BF16): 10/20/2026 via BF16 Model Card
Hugging Face (NVFP4): 10/20/2026 via NVFP4 Model Card

Model Architecture:

Architecture Type: Transformer
Number of Model Parameters: 1.55B

BF16

Network Architecture: Nemotron-3.5-Embed-1B-BF16 consists of a Pixtral vision encoder initialized from mistralai/Ministral-3-3B-Instruct-2512-BF16 and a language model initialized from nvidia/Nemotron-3-Embed-1B-BF16.

The Nemotron-3.5-Embed-1B-BF16 model was distilled from the Nemotron-3.5-Embed-8B-BF16 embedding teacher using combined cosine distance (COS) and mean squared error (MSE) losses on multilingual text and image retrieval data. Its language-model component was initialized from Nemotron-3-Embed-1B-BF16 and kept frozen during distillation. Consequently, Nemotron-3.5-Embed-1B-BF16 shares the same text embedding space as Nemotron-3-Embed-1B-BF16 while adding image retrieval capabilities. The vision encoder and multimodal projection were fine-tuned on image and combined image-text datasets.

NVFP4

Network Architecture: The Nemotron-3.5-Embed-1B-NVFP4 is the quantized version of the Nemotron-3.5-Embed-1B-BF16, which is a transformer-based multimodal embedding model distilled from a fine-tuned Nemotron-3.5-Embed-8B-BF16 teacher model.

For distribution, the NVFP4 checkpoint was converted to the compressed-tensors format. NVFP4 quantization is applied to eligible linear layers in both the language model and the vision encoder. The final embedding vector is obtained by applying average pooling to the transformer’s token-level representations. It encodes each input (image, text, or image + text) into a dense embedding vector of dimension 2048.

Post-Training Quantization and Quantization-Aware Distillation

This model is a post-training quantized variant of Nemotron-3.5-Embed-1B-BF16. Quantization was applied to the weights and activations of linear layers only, targeting the NVFP4 data type for efficient inference. Quantization-Aware Distillation (QAD) was applied primarily to recover accuracy for long input sequences.

Input:

Input Type(s): Text, Image
Input Format(s): Text: String; Image: Red, Green, Blue (RGB)
Input Parameters: Text: One-Dimensional (1D), Image: Two-Dimensional (2D)
Other Properties Related to Input: The model supports image and/or text inputs. For queries, the model accepts text input only. For documents, it accepts an image, text, or a combination of image + text. The default image longest_edge is 1,120 pixels, and it can be increased up to 1,540 pixels. The source models’ max context length is 32,768 tokens.
Hosted API Input Handling: Image inputs use input_type="passage"; queries are text-only. truncate="END" removes text from the end, while truncate="NONE" returns an error for over-limit inputs. An image block that alone exceeds the limit is rejected. See the API reference.

Output:

Output Type(s): Embeddings
Output Format: Vector
Output Parameters: One-Dimensional (1D) per input
Other Properties Related to Output: Token-level hidden states are mean-pooled and then L2-normalized to produce one 2048-dimensional embedding vector for each input. The model also supports dynamic embedding sizes by truncating the vector from the end (e.g., retaining the first 1024 or 512 dimensions). The truncated embeddings remain effective for retrieval, provided they are L2-normalized after truncation.

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA's hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.

Software Integration:

Runtime: NVIDIA hosted Preview API
API Endpoint and Request/Response Format: POST https://integrate.api.nvidia.com/v1/embeddings with JSON requests and embedding responses; model ID nvidia/nemotron-3.5-embed-1b. See the API reference.

The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.

Model Version(s):

• Nemotron-3.5-Embed-1B-BF16: The variant served by the NVIDIA API Catalog Preview API. See the BF16 Model Card.
• Nemotron-3.5-Embed-1B-NVFP4: The NVFP4 quantized derivative, covered here with its own architecture, quantization, training, runtime, and evaluation details. See the NVFP4 Model Card.

Shared input/output properties and the four Model Card++ subcards apply to both variants. Variant-specific disclosures are identified explicitly.

Training, Testing, and Evaluation Datasets:

BF16 Training Dataset

Data Modality:

• Image
• Text

Image Training Data Size: 2.3M samples containing images, with or without accompanying text
Text Training Data Size: 8.5M+ text-only data samples
Data Collection Method by dataset: Automated, Manually-Labeled, Synthetic
Labeling Method by dataset: Automated, Manually-Labeled, Synthetic
Properties: Training was conducted using publicly available, commercially permissible datasets and synthetically generated datasets.

Public Datasets:

The language model of Nemotron-3.5-Embed-1B-BF16 was trained with text-only datasets that have been listed in Nemotron-3-Embed-1B-BF16's Model Card. In addition to these datasets we have used the following publicly available datasets to train the vision encoder and multimodal projector with image-only and image + text inputs.

Synthetic datasets were created either by generating queries from seed documents or by generating complete question–answer pairs through LLM-based prompting using the models listed below.

NVFP4 Quantization and QAD Dataset

Data Modality:

• Image
• Text

Image Training Data Size: 49k samples
Text Training Data Size: 49k samples
Data Collection Method by dataset: Automated, Manually-Labeled, Synthetic
Labeling Method by dataset: Automated, Manually-Labeled, Synthetic
Properties: This checkpoint is an NVFP4 post-training-quantized derivative of nvidia/Nemotron-3.5-Embed-1B-BF16. NVFP4 quantization and Quantization-Aware Distillation (QAD) used a 49k-sample subset of the training data, selected to cover typical cases and tail cases, and containing both images and text. Of these samples, initial post-training quantization calibration used 384 samples: 128 images, 128 short texts, and 128 long texts. Initial QAD of the vision encoder used 5,120 randomly sampled images. Later text and image+text QAD stages used the full 49k-sample subset, with 5% held out for validation.

Nemotron-3.5-Embed-1B-BF16's training was conducted using publicly available, commercially permissible datasets and synthetically generated datasets. For more detail about the text and image datasets used for training please refer to the BF16 model card.

Testing Dataset

Data Collection Method by dataset: Not Applicable
Labeling Method by dataset: Not Applicable
Properties: Not Applicable. Model quality was assessed using the evaluation benchmark datasets described in the Evaluation Dataset subsection.

BF16 Evaluation Dataset

Benchmark:

• ViDoRe V3
• MIRACL-VISION
  

Benchmark Score: NDCG@10

The Nemotron-3.5-Embed-1B-BF16 and Nemotron-3.5-Embed-8B-BF16 were evaluated with the models' default image longest_edge of 1,120 pixels. Qwen3-VL-Embedding-2B and Qwen3-VL-Embedding-8B were evaluated with max_pixels set to 1,843,200.

Detailed BF16 Benchmark Comparison

Model	ViDoRe V3			MIRACL-VISION
	image	text	image + text	image
Qwen3-VL-Embedding-2B	52.89	48.31	51.97	56.19
Nemotron-3.5-Embed-1B-BF16	58.02	57.80	58.82	66.56
Qwen3-VL-Embedding-8B	58.29	53.69	57.03	65.87
Nemotron-3.5-Embed-8B-BF16	60.39	60.65	61.47	70.15

Data Collection Method by dataset: Automated, Manually-Labeled, Synthetic
Labeling Method by dataset: Automated, Manually-Labeled, Synthetic
Properties: ViDoRe V3 is a visual document retrieval benchmark covering 8 public and 2 private datasets across professional domains. MIRACL-VISION is a multilingual visual retrieval dataset for 18 different languages. It is an extension of MIRACL, a popular text-only multilingual retrieval dataset.

NVFP4 Evaluation Dataset

Benchmark:

• ViDoRe V3
  

Benchmark Score: NDCG@10

Both Nemotron-3.5-Embed-1B-BF16 and Nemotron-3.5-Embed-1B-NVFP4 were evaluated with the models' default image longest_edge of 1,120 pixels.

Data Collection Method by dataset: Automated, Manually-Labeled, Synthetic
Labeling Method by dataset: Automated, Manually-Labeled, Synthetic
Properties: ViDoRe V3 is a visual document retrieval benchmark covering 8 public and 2 private datasets across professional domains.

Inference:

Served Variant: BF16
Acceleration Engine: Other (Photon Engine; native CUDA backend)
Test Hardware: NVIDIA Ampere (A100, A10G), NVIDIA Blackwell (GB10, GB200, RTX-PRO-6000-BSE), NVIDIA Hopper (H100, H100-NVL), NVIDIA Lovelace (L40S)

Source Model Inference Requirements

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. Developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please make sure you have proper rights and permissions for all input image and video content; if image or video includes people, personal health information, or intellectual property, the image or video generated will not blur or maintain proportions of image subjects included.

For more detailed information on ethical considerations for both variants, see the Model Card++ documents:

• Bias
• Explainability
• Privacy
• Safety & Security

Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.