Nemotron-3-Embed-1B-NVFP4 is the quantized version of the Nemotron-3-Embed-1B-BF16 model, which is developed for text question-answering retrieval. For more information, please check here. The Nemotron-3-Embed-1B-NVFP4 model is quantized with NVIDIA Model Optimizer, using nvidia-modelopt v0.45.0.
This model was evaluated on 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, Vietnamese. Read more details in our Blog Post.
This project will download and install additional third-party open source software projects. Review the license terms of these open source projects before use.
Deployment Geography
Global
Use Case
The Nemotron-3-Embed-1B-NVFP4 is most suitable for users who want to build a multilingual question-and-answer application over a large text corpus, leveraging the latest dense retrieval technologies.
Generally Nemotron-3-Embed-1B-BF16 and Nemotron-3-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.
Number of model parameters: The model has 1.14B parameters
Hidden Size: 2048
The Nemotron-3-Embed-1B-NVFP4 is the quantized version of the Nemotron-3-Embed-1B-BF16, which is a transformer-based text embedding model trained with bidirectional attention masking, where the final embedding vector is obtained by applying average pooling to the transformer’s token-level representations. It encodes each input text into a dense embedding vector of dimension 2048.
Input(s)
Input Type(s): Text
Input Format(s):
Text: List of strings
Input Parameters:
Text: One-Dimensional (1D)
Other Properties Related to Input: Text inputs should be tokenized by the model tokenizer. The model’s max sequence length is 32768. Longer inputs should be chunked or truncated.
Output(s)
Output Type(s): Floats
Output Format(s):
List of float arrays
Output Parameters: One-Dimensional (1D) embedding vector per input text string
Other Properties Related to Output: The model outputs a 2048-dimensional embedding vector for each input text string. It also supports dynamic embedding sizes by slicing the vector from the start (for example, keeping the first 1024 or 512 dimensions). These sliced embeddings remain highly functional, provided the resulting sub-vector is re-normalized (L2 normalization) after slicing.
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.
Post-Training Quantization and Quantization-Aware Distillation
This model is a post-training quantized variant of nvidia/Nemotron-3-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.
vLLM Usage
This NVFP4 checkpoint is intended for vLLM. The examples below show offline Python use and online serving. Use nvidia/Nemotron-3-Embed-1B-BF16 if you need Transformers or Sentence Transformers.
Use the Hugging Face model ID by default. If you are working from a local checkpoint, replace MODEL_ID with that path:
MODEL_ID=nvidia/Nemotron-3-Embed-1B-NVFP4
The output tables use q[i] for queries and d[i] for documents. Scores are rounded to four decimal places, and runtime differences might affect the final decimal places.
Tested vLLM Versions
This checkpoint has been explicitly tested with the following vLLM distributions:
Use vLLM 0.25.0 for the examples below. vLLM 0.23.x and 0.24.x have known issues with this NVFP4 checkpoint family. Other versions have not been explicitly validated.
To install vLLM without a container image, run the following command:
Use the offline Python API for local vLLM inference without an HTTP server. LLM.embed accepts formatted strings. Add the query: and passage: prefixes manually. The example uses a 4,096-token limit. Review CUDA graph sizing before increasing it.
vLLM Offline Python Example
python
1import numpy as np
2from vllm import LLM
34MODEL_ID ="nvidia/Nemotron-3-Embed-1B-NVFP4"5MAX_MODEL_LEN =40966MAX_BATCHED_TOKENS =409678QUERIES =[9"Write a Python function that counts the frequency of each element in a list of lists.",10"Write a function that orders a dictionary with tuple keys by the product of each key's tuple values.",11"What symptoms and common triggers help distinguish eczema from other inflammatory skin conditions?",12"How can someone reduce exposure to pollen during allergy season?",13]1415DOCUMENTS =[16"def frequency_lists(list1):\n flattened = [item for sublist in list1 for item in sublist]\n counts = {}\n for item in flattened:\n if item in counts:\n counts[item] += 1\n else:\n counts[item] = 1\n return counts",17"def sort_dict_item(test_dict):\n return {key: test_dict[key] for key in sorted(test_dict.keys(), key=lambda ele: ele[0] * ele[1])}",18"Eczema commonly causes itchy, dry, inflamed patches of skin. The affected areas may look red, scaly, cracked, or darker than the surrounding skin depending on skin tone. Symptoms can flare after exposure to irritants, allergens, stress, or changes in weather.",19"People with pollen allergy can reduce exposure by staying indoors on dry, windy days, avoiding early-morning outdoor activity, and going outside after rain when pollen levels are lower. They should check pollen forecasts, close windows and doors when counts are high, and consider starting allergy medication before symptoms begin if high pollen is expected. After being outside, showering, changing clothes, avoiding outdoor laundry drying, and wearing a face mask for yard work can help limit pollen contact.",20]2122defmain():23 llm = LLM(24 model=MODEL_ID,25 max_model_len=MAX_MODEL_LEN,26 max_num_batched_tokens=MAX_BATCHED_TOKENS,27 max_cudagraph_capture_size=MAX_BATCHED_TOKENS,28)29 texts =["query: "+ query for query in QUERIES]+[30"passage: "+ doc for doc in DOCUMENTS
31]32 outputs = llm.embed(texts, use_tqdm=False)33 embeddings = np.array(34[output.outputs.embedding for output in outputs],35 dtype=np.float32,36)3738 query_embeddings = embeddings[:len(QUERIES)]39 document_embeddings = embeddings[len(QUERIES):]4041 scores = query_embeddings @ document_embeddings.T
42print("Similarity scores:")43print(f"{'':>8}"+"".join(f"d[{i}] "for i inrange(scores.shape[1])))44for query_index, row inenumerate(scores):45print(f"q[{query_index}] "+" ".join(f"{score:>7.4f}"for score in row))464748if __name__ =="__main__":49 main()
The checkpoint supports sequences up to 32,768 tokens. The examples use 4,096 as a conservative starting point.
Use the following guidance to tune CUDA graph capture:
Set --max-model-len to the longest request you intend to serve. Tune --max-num-batched-tokens for the workload, concurrency, and available GPU memory. When chunked prefill is disabled, the batched-token budget must be at least the model-length limit.
For default capture buckets up to 8,192, set --max-cudagraph-capture-size equal to --max-num-batched-tokens. This setting makes batches up to the scheduler budget eligible for CUDA graph execution. Batches outside the captured range use a slower uncaptured path.
Larger capture ranges increase startup time and graph memory. Benchmark representative traffic on the target hardware. Do not capture beyond the batched-token budget.
Use a maximum capture size of 4,096 as the conservative default for services that restart or autoscale regularly. A maximum capture size of 8,192 can be reasonable when a longer cold start is acceptable. For maximum capture sizes above 8,192, pass a smaller, workload-aligned set with --cudagraph-capture-sizes to keep startup time under control.
Example Sparse Capture Sizes for Inputs up to 32,768 Tokens
The following command uses a sparse capture-size list:
Choose sizes from representative batch-token measurements. vLLM pads each execution batch to the next captured size, so denser lists reduce padding but require more startup time and graph memory.
In illustrative tests on an NVIDIA GB10 system with vLLM 0.25.0, cold startup using automatic buckets took about 74 seconds at a maximum capture size of 4,096 and 121 seconds at 8,192. At 32,768, startup with automatic buckets was projected to take tens of minutes. With the sparse list of 49 capture sizes above, startup completed in about one minute. Results vary by workload and hardware.
Add --host or --port to the serving command if you need non-default network settings.
To serve a local checkpoint, replace MODEL_ID with its path. Add --served-model-name nvidia/Nemotron-3-Embed-1B-NVFP4 if clients should continue using the Hugging Face model ID.
Recommended Retrieval Endpoint
After the server is running, use /v2/embed for retrieval. Send raw query and document strings. input_type applies the saved query and document prompt metadata.
The following example uses the recommended endpoint:
python
1import numpy as np
2import requests
34MODEL ="nvidia/Nemotron-3-Embed-1B-NVFP4"5URL ="http://localhost:8000/v2/embed"67QUERIES =[8"Write a Python function that counts the frequency of each element in a list of lists.",9"Write a function that orders a dictionary with tuple keys by the product of each key's tuple values.",10"What symptoms and common triggers help distinguish eczema from other inflammatory skin conditions?",11"How can someone reduce exposure to pollen during allergy season?",12]1314DOCUMENTS =[15"def frequency_lists(list1):\n flattened = [item for sublist in list1 for item in sublist]\n counts = {}\n for item in flattened:\n if item in counts:\n counts[item] += 1\n else:\n counts[item] = 1\n return counts",16"def sort_dict_item(test_dict):\n return {key: test_dict[key] for key in sorted(test_dict.keys(), key=lambda ele: ele[0] * ele[1])}",17"Eczema commonly causes itchy, dry, inflamed patches of skin. The affected areas may look red, scaly, cracked, or darker than the surrounding skin depending on skin tone. Symptoms can flare after exposure to irritants, allergens, stress, or changes in weather.",18"People with pollen allergy can reduce exposure by staying indoors on dry, windy days, avoiding early-morning outdoor activity, and going outside after rain when pollen levels are lower. They should check pollen forecasts, close windows and doors when counts are high, and consider starting allergy medication before symptoms begin if high pollen is expected. After being outside, showering, changing clothes, avoiding outdoor laundry drying, and wearing a face mask for yard work can help limit pollen contact.",19]2021defembed(input_type:str, texts:list[str])-> np.ndarray:22 response = requests.post(23 URL,24 json={25"model": MODEL,26"input_type": input_type,27"texts": texts,28"embedding_types":["float"],29"truncate":"END",30},31 timeout=120,32)33 response.raise_for_status()34return np.array(response.json()["embeddings"]["float"], dtype=np.float32)3536query_embeddings = embed("query", QUERIES)37document_embeddings = embed("document", DOCUMENTS)3839scores = query_embeddings @ document_embeddings.T
40print("Similarity scores:")41print(f"{'':>8}"+"".join(f"d[{i}] "for i inrange(scores.shape[1])))42for query_index, row inenumerate(scores):43print(f"q[{query_index}] "+" ".join(f"{score:>7.4f}"for score in row))
You can also use the OpenAI-compatible /v1/embeddings endpoint. For those requests, pass strings in input and manually prefix them with query: or passage: .
Expected Configuration Warning
When vLLM loads this checkpoint, its Transformers configuration parser can emit the following warning.
[transformers] Unrecognized keys in `rope_parameters` for 'rope_type'='yarn': {'apply_yarn_scaling'}
This warning is expected and does not prevent vLLM from loading the model or running inference. apply_yarn_scaling is a temporary vLLM compatibility field that preserves the checkpoint's intended long-context rotary position embedding (RoPE) behavior. Do not remove it from config.json. Refer to vLLM issue #48621 for upstream compatibility work.
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-Embed-1B-NVFP4
Training, Testing, and Evaluation Datasets
Dataset Overview
This checkpoint is an NVFP4 post-training-quantized derivative of nvidia/Nemotron-3-Embed-1B-BF16. Quantization-Aware Distillation (QAD) was applied on a small sample of the original BF16 data mix for further accuracy recovery.
Calibration Dataset
Data Collection Method by dataset: Automated
Labeling Method by dataset: Automated
Properties: A calibration dataset of 512 total samples was used for NVFP4 post-training quantization, consisting of 256 queries and 256 passages from the abisee/cnn_dailymail dataset, formatted with query and passage prefixes.
Training Dataset
QAD training was performed as part of the NVFP4 model development. The information below describes the QAD training datasets used for this model. For details about the training datasets used to build the underlying base model, Nemotron-3-Embed-1B-BF16, please refer to its model card.
Total Size: 20k data samples
Total Number of Datasets: 5 dataset files
Dataset Partition: Training [100%], Testing [N/A — evaluation benchmarks used separately], Validation [N/A — evaluation benchmarks used separately].
Data Collection Method by dataset: Hybrid: Human, Automated, Synthetic
Labeling Method by dataset: Hybrid: Human, Automated, Synthetic
Properties: Model distillation training of the underlying model was conducted on text datasets using question–passage pairs from publicly available, commercially permissible datasets and synthetically generated datasets. For more information, please visit this link.
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.
Evaluation Dataset
Data Collection Method by dataset: Hybrid: Human, Automated, Synthetic
Labeling Method by dataset: Hybrid: Human, Automated, Synthetic
Properties: In this section, we compare the performance of quantized model Nemotron-3-Embed-1B-NVFP4 with baseline implementation Nemotron-3-Embed-1B-BF16.
This model is evaluated on 16 public tasks on Retrieval Embedding Benchmark (RTEB), a new benchmark designed to reliably evaluate the retrieval accuracy of embedding models for real-world applications. More details on RTEB can be found on their leaderboard.
We set the model sequence length to 4096 for the evaluation results below. The NVFP4 model was evaluated on an NVIDIA GB200 GPU.
Text Retrieval Benchmarks (chunk retrieval) – Avg. NDCG@10
Model Name
Precision
RTEB
Nemotron-3-Embed-1B-BF16
BF16
72.38
Nemotron-3-Embed-1B-NVFP4
NVFP4
72.00
Inference
Acceleration Engine: vLLM
Test Hardware:
NVIDIA Ampere - A100 PCIe/SXM
NVIDIA Blackwell - GB200 and RTX 6000 PRO
NVIDIA Hopper - H100 PCIe/SXM
NVIDIA Lovelace - L40 and L4
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.
For more detailed information on ethical considerations for this model, please see the Model Card++ Bias, Explainability, Safety & Security, and Privacy Subcards.
Please report model quality, risk, security vulnerabilities or NVIDIA AI concerns here.
Bias
Field
Response
Participation considerations from adversely impacted groups protected classes in model design and testing
None
Measures taken to mitigate against unwanted bias
None
Bias Metric (If Measured):
None
Explainability
Field
Response
Intended Task/Domain:
Passage and query embedding for question and answer retrieval
Model Type:
Transformer encoder
Intended Users:
Generative AI creators working with conversational AI models - users who want to build a multilingual question and answer application over a large text corpus, leveraging the latest dense retrieval technologies.
Output:
Array of float numbers (Dense Vector Representation for the input text)
Describe how the model works:
Model transforms the tokenized input text into a dense vector representation.
Name the adversely impacted groups this has been tested to deliver comparable outcomes regardless of:
Not Applicable
Technical Limitations & Mitigation:
The model’s max sequence length is 32768. Therefore, the longer text inputs should be truncated.
Verified to have met prescribed NVIDIA quality standards:
Yes
Performance Metrics:
Accuracy, Throughput, and Latency
Potential Known Risks:
This model does not always guarantee to retrieve the correct passage(s) for a given query.
The Principle of least privilege (PoLP) is applied limiting access for dataset generation and model development. Restrictions enforce dataset access during training, and dataset license constraints adhered to.
