Nemotron 3.5 Parse
Description
Nemotron 3.5 Parse is a document image parser that produces structured, machine-readable text with semantic layout classes, bounding boxes, and reading order. It combines a C-RADIOv4-SO400M vision encoder with a 10-layer mBART decoder that consumes the visual sequence as a causal prefix. The release uses dynamic rectangular NaViT preprocessing, content-normalized coordinates, grouped-query self-attention (16 query heads and 8 key/value heads), a 72,505-token multilingual tokenizer, and chart-aware output through the <class_Chart> token. It is intended for document understanding, information retrieval, data extraction, and multimodal data-curation workflows.
Nemotron 3.5 Parse was developed by NVIDIA as a part of Nemotron.
This model is ready for commercial or non-commercial use.
License and Terms of Use:
GOVERNING TERMS: Use of this trial service is governed by the NVIDIA API Trial Terms of Service. Use of the model is governed by the OpenMDW License Agreement, version 1.1.
Model and Tokenizer Terms: Your use of this model is governed by the OpenMDW License Agreement, version 1.1. Use of the tokenizer included in this model is governed by the CC-BY-4.0 license. See third-party notices for separately installed dependencies.
Deployment Geography:
Global
Use Case:
Use Case: Nemotron 3.5 Parse is designed for developers and teams building document intelligence, retrieval-augmented generation (RAG), curator, extractor, and agentic AI applications. It can convert scanned or rendered PDFs, presentation slides, forms, reports, tables, and mixed-content document pages into structured outputs for downstream indexing, retrieval, analytics, model training-data creation, and human-in-the-loop review.
Release Date:
Build.NVIDIA.com: 10/20/2026 via link
Hugging Face: 10/20/2026 via link
Reference(s):
References:
Model Architecture:
Architecture Type: Transformer
Network Architecture:
- Vision Encoder: NVIDIA C-RADIOv4-SO400M (
vit_so400m_patch16_224) with dynamic rectangular NaViT preprocessing - Adapter Layer: 1D convolutions and normalization layers that compress the vision latent sequence before decoding
- Decoder: 10-layer mBART causal visual-prefix decoder with 16 query heads and 8 key/value heads
- Auxiliary Prediction Head: One trained one-step multi-token-prediction (MTP) head is included in
model.safetensors. Transformers uses the normal decoder for generation; the vLLM plugin enables the bundled head for one-token speculative decoding. - Tokenizer: 72,505 entries with multilingual and task/control tokens including
<predict_bbox>,<predict_classes>,<predict_text_in_pic>, and<class_Chart>; the decoder embedding is padded to 72,512 rows. Use of the tokenizer included in this model is governed by the CC-BY-4.0 license.
Total Parameters: 630.8M (630,800,496 unique trainable parameters, including the MTP head; shared decoder embedding/output weights are counted once)
Input:
Input Types: Image, Text
Input Formats:
- Image: Red, Green, Blue (RGB)
- Text: Prompt string
Input Parameters: Image: Two-Dimensional (2D), Text: One-Dimensional (1D)
Other Input Properties:
- Recommended maximum input resolution (Width, Height): 1664, 2048
- Channel Count: 3
- Prompt format: one task prompt composed from the supported control tokens. The default prompt extracts bounding boxes, semantic classes, and text in markdown format:
</s><s><predict_bbox><predict_classes><output_markdown><predict_text_in_pic>.
Output:
Output Types: Text
Output Format: String
Output Parameters: One-Dimensional (1D)
Other Output Properties: Nemotron 3.5 Parse returns a string encoding document text, semantic element classes, and content-normalized bounding boxes. Scale x coordinates by the evaluated image width and y coordinates by its height. Postprocessing utilities can convert table or chart-associated text into LaTeX, HTML, markdown, JSON, or CSV where supported.
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 Engine(s):
- Native CUDA backend: Photon Engine
Supported Hardware Microarchitecture Compatibility:
- NVIDIA Ampere: A100, A10G
- NVIDIA Blackwell: GB200, GB10, RTX PRO 6000 Blackwell Server Edition
- NVIDIA Hopper: H100 NVL
- NVIDIA Lovelace: L40S
Supported Operating System(s): Linux
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 Parse
Training, Testing, and Evaluation Datasets:
Training Dataset
Data Modality: Image, Text
Image Training Data Size: 1 Million to 1 Billion Images
Text Training Data Size: 1 Billion to 10 Trillion Tokens
Data Collection Method by dataset: Hybrid: Automated, Manually-Collected, Synthetic
Labeling Method by dataset: Hybrid: Automated, Manually-Labeled, Synthetic
Labeling Details: Labels are produced through a combination of automated OCR and layout annotation, manual annotation, and synthetic generation. The combination varies by constituent dataset.
Properties: The training set contains millions of image-text items aggregated across large document, table, and layout datasets. The data consists of document-page and table images paired with OCR text, bounding boxes, and layout labels. Sources include rendered digital documents, scientific papers, PDFs, Wikipedia-style pages, and synthetic document, table, word, and character renderings. Annotations come from OCR and layout models, third-party OCR services, synthetic-generation pipelines, and human labeling.
Testing Dataset
Data Collection Method by dataset: Hybrid: Automated, Manually-Collected, Synthetic
Collection Details: Sources combine automated collection, manually collected material, and synthetic generation; methods vary by constituent dataset.
Labeling Method by dataset: Hybrid: Automated, Manually-Labeled, Synthetic
Labeling Details: Labels are produced through a combination of automated OCR and layout annotation, manual annotation, and synthetic generation. The combination varies by constituent dataset.
Properties: Tested on internal and public document-understanding benchmarks using document-page images and reference text, layout classes, bounding boxes, and structure annotations. Coverage includes OCR quality, layout structure, table parsing, reading order, and visual grounding across document types and languages.
Evaluation Dataset
Data Collection Method by dataset: Hybrid: Automated, Manually-Collected, Synthetic
Collection Details: Sources combine automated collection, manually collected material, and synthetic generation; methods vary by constituent dataset.
Labeling Method by dataset: Hybrid: Automated, Manually-Labeled, Synthetic
Labeling Details: Labels are produced through a combination of automated OCR and layout annotation, manual annotation, and synthetic generation. The combination varies by constituent dataset.
Properties: Evaluated on internal and public document-understanding benchmarks covering OCR quality, layout structure, table parsing, reading order, and visual grounding. The ParseBench reference evaluation covers 2,078 unique document pages. Additional benchmarks cover Indic-language OCR and multilingual synthetic OCR across the scripts listed below.
Benchmark coverage:
- ParseBench evaluates document parsing across text fidelity, semantic formatting, tables, charts, and visual grounding with layout boxes/classes.
- IndicVisionBench evaluates OCR quality for Indic-language pages, including Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, and Telugu.
- MOSCAR evaluates multilingual synthetic OCR across broad script coverage, including Latin, Arabic, Cyrillic, Han, Hangul, Japanese, Indic scripts, Hebrew, Thai, Greek, and others.
Evaluation Results
Evaluation Benchmark Score: ParseBench mean 69.05; MOSCAR corpus CER 1.46; IndicVision character ANLS 82.00.
#1 ParseBench among models below 1B parameters, with 4.6× the measured page throughput of PaddleOCR-VL 1.6. Nemotron 3.5 Parse scores 69.05 and processes 14.21 pages/s in the reference BF16 profile.
Detailed Benchmark Comparison and Methodology
| Model | Mean ↑ | Tables ↑ | Charts ↑ | Content Faithfulness ↑ | Semantic Formatting ↑ | Visual Grounding ↑ | Pages/s ↑ |
|---|---|---|---|---|---|---|---|
| PaddleOCR-VL 1.6 | 67.43 | 67.77 | 54.24 | 82.71 | 54.64 | 77.80 | 3.09 |
| Nemotron Parse 2.0 | 64.46 | 70.41 | 31.59 | 84.72 | 65.36 | 70.23 | 3.97 |
| Nemotron 3.5 Parse | 69.05 | 72.60 | 49.89 | 86.83 | 65.56 | 70.26 | 14.21 |
Paddle accuracy uses its unchanged official published results; its throughput uses the optimized full native pipeline on the same 2,078 pages and one H100. Paddle retains a 16,384-token context and 8,192-token output cap. Parse uses one-token MTP, a 16,384-token context, and an 8,744-token output cap. Timing excludes startup, image preload, and 16 warmups. Paddle timing and settings are included.
| Benchmark / metric | Direction | Nemotron Parse 2.0 | Nemotron 3.5 Parse |
|---|---|---|---|
| MOSCAR · corpus CER | ↓ | 4.25 | 1.46 |
| IndicVision · character ANLS | ↑ | 67.44 | 82.00 |
Inference reference outputs and verification instructions are available in reference/.
Inference
Acceleration Engine: Native CUDA backend: Photon Engine
Test Hardware:
- NVIDIA Ampere (A100)
- NVIDIA Ampere (A10G)
- NVIDIA Blackwell (GB200)
- NVIDIA Blackwell (GB10)
- NVIDIA Blackwell (RTX PRO 6000 Blackwell Server Edition)
- NVIDIA Hopper (H100 NVL)
- NVIDIA Lovelace (L40S)
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 an input image contains people, personal health information, confidential business data, or intellectual property, the model may extract or reproduce visible text from that content.
For more detailed information on ethical considerations for this model, please see the Explainability, Bias, Safety, and Privacy subcards. Please report model quality, risk, security vulnerabilities, or NVIDIA AI Concerns here.