[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-657504-105":59,"doc-detail-657504-en":129},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":122,"head_meta":124,"extra_data":126,"updated_unix":128},105,"en","nemotron-parse-11-lightweight-document-parsing-and-ocr-model","Nemotron-Parse 1.1 - Lightweight Document Parsing and OCR Model","","Nemotron-Parse-1.1 is a lightweight document parsing and OCR model that improves general OCR, Markdown formatting, structured table parsing, and text extraction from pictures, charts, and diagrams. It preserves bounding boxes for text segments and corresponding semantic classes, supporting longer outputs for visually dense documents. Built on an encoder-decoder design with 885M parameters (including a 256M language decoder), it releases model weights and a faster token-compressed variant with about 20% speedup.",{"@graph":69,"@context":121},[70,84,104],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":30,"@type":76,"position":81},"https://docshare.wps.com/document/technology/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/nemotron-parse-11-lightweight-document-parsing-and-ocr-model/657504/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":98,"encodingFormat":97,"isAccessibleForFree":99,"interactionStatistic":100},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/nemotron-parse-11-lightweight-document-parsing-and-ocr-model/657504.png","ImageObject",300,407,{"name":92,"@type":93},"Ivy","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-26",true,{"@type":101,"interactionType":102,"userInteractionCount":8},"InteractionCounter",{"@type":103},"ViewAction",{"@type":105,"mainEntity":106},"FAQPage",[107,113,117],{"name":108,"@type":109,"acceptedAnswer":110},"What capabilities does Nemotron-Parse-1.1 improve compared with its predecessor?","Question",{"text":111,"@type":112},"It improves general OCR, Markdown formatting, structured table parsing, and text extraction from pictures, charts, and diagrams. It also supports longer output sequences for visually dense documents.","Answer",{"name":114,"@type":109,"acceptedAnswer":115},"What outputs does Nemotron-Parse-1.1 provide besides extracted text?",{"text":116,"@type":112},"It extracts bounding boxes of text segments and provides corresponding semantic classes, while preserving the reading order.",{"name":118,"@type":109,"acceptedAnswer":119},"How does the token-compressed variant (Nemotron-Parse-1.1-TC) differ from the full model?",{"text":120,"@type":112},"Nemotron-Parse-1.1-TC uses a reduced vision token length to reduce latency and computation. It offers about a 20% speed improvement with minimal quality degradation and maintains core document understanding features.","https://schema.org",{"og:url":83,"og:type":123,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":125,"canonical":83},"index,follow",{"doc_id":127,"site_id":62},657504,1793051710,{"code":4,"msg":5,"data":130},{"doc_id":127,"user_id":131,"nickname":92,"user_avatar":132,"doc_module":4,"category_id":29,"category_name":30,"doc_title":65,"doc_description":67,"doc_content":133,"file_id":134,"file_url":135,"file_type":136,"file_size":137,"view_count":8,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":138,"language":139,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":67,"update_tm":143,"read_time":144},549758252649,"https://ap-avatar.wpscdn.com/avatar/8000253669c5317157?_k=1778319167496531819","arXiv :2511 .20478v1 [ cs .LG] 25 Nov 2025  \nNVIDIA Nemotron-Parse 1.1  \nNVIDIA  \nAbstract. We introduce Nemotron-Parse-1 . 1, a lightweight document parsing and OCR model that advances the capabilities of its predecessor, Nemoretriever-Parse-1.0 (Karmanov et al. , 2025) . Nemotron-Parse-1.1 delivers improved capabilities across general OCR, markdown formatting, structured table parsing, and text extraction from pictures, charts, and diagrams. It also supports a longer output sequence length for visually dense documents. As with its predecessor, it extracts bounding boxes of text segments, as well as corresponding semantic classes.  \nNemotron-Parse-1.1 follows an encoder-decoder architecture with 885M parameters, including a compact 256M-parameter language decoder. It achieves competitive accuracy on public benchmarks making it a strong lightweight OCR solution. We release the model weights publicly on Huggingface, as well as an optimized NIM container, along with a subset of the training data as part of the broader Nemotron-VLM-v2 dataset. Additionally, we release Nemotron-Parse-1 . 1-TC which operates on a reduced vision token length, offering a 20% speed improvement with minimal quality degradation.  \n1. Introduction  \nIn the recent years, document-level OCR has developed beyond simple extraction of plain text characters from an image. Modern applications such as Large Language Models, retrieval systems, question-answering solutions, demand a richer representation, which includes layout, reading order, semantic classes (such as captions or footnotes), formulas, tables, and understanding of multicolumn/multi-page structure.  \nA number of models were introduced in recent years, ranging from pipeline-based solutions to end-to-end Vision-Language models (Feng et al. , 2025 ; Marker, 2025 ; Mathpix, 2025 ; Wang et al. , 2024 ; Li et al. , 2025 ; Cui et al. , 2025 ; Blecher et al. , 2024 ; Nassar et al. , 2022 , 2025 ; Poznanski et al. , 2025 ; Wei et al. , 2024 ; ocr, 2025 ; Rednote) .  \nPipeline solutions often rely on brittle multi-stage pipelines with each stage responsible for a subtask of the overall document extraction process, hence achieving versatility at the cost of the lower throughput. At the same time, end-to-end models benefit from fast inference speeds while often not performing equally well on all subtasks associated with document extraction simultaneously, i.e. , general OCR, structured text formatting, mathematical equations, extraction of complex tables, prediction of bounding boxes and semantic classes of text blocks.  \nTo close this gap, we introduce Nemotron Parse 1.1, a successor to Nemoretriever-Parse 1.0 and Eclair (Karmanov et al. , 2025), which is an end-to-end vision-language model capable of extracting formatted text (Markdown/LaTeX), bounding boxes of text blocks, and semantic classes for each block while preserving the reading order. Beyond the full-capability Nemotron Parse 1.1 model, we also introduce Nemotron Parse 1.1 with token compression (TC), a streamlined variant designed for applications that prioritize speed without majorly sacrificing output quality. Nemotron-ParseTC maintains nearly the same level of accuracy as the full model, while offering reduced latency and lower computational overhead. This makes it particularly well-suited for large-scale batch processing, edge deployments, or interactive systems where rapid response times are critical. Despite  \nits lightweight architecture, Nemotron-Parse-TC preserves support for core document understanding features—including layout-aware text extraction, semantic block classification, and consistent readingorder generation—providing an efficient alternative for production environments that demand both performance and precision. Additionally, Nemotron-Parse-TC offers an improved reading order, with all page elements, including floating ones, following the page ordering.  \n2. Nemotron-Parse  \n2.1. Model Architecture  \nNemotron-parse follows an ","cbCaiiXrloIu05PC","https://ap.wps.com/l/cbCaiiXrloIu05PC","pdf",2701095,14,"English","# Introduction\n## Nemotron-Parse\n## Model Architecture\n## Positional embeddings","[{\"question\":\"What capabilities does Nemotron-Parse-1.1 improve compared with its predecessor?\",\"answer\":\"It improves general OCR, Markdown formatting, structured table parsing, and text extraction from pictures, charts, and diagrams. It also supports longer output sequences for visually dense documents.\"},{\"question\":\"What outputs does Nemotron-Parse-1.1 provide besides extracted text?\",\"answer\":\"It extracts bounding boxes of text segments and provides corresponding semantic classes, while preserving the reading order.\"},{\"question\":\"How does the token-compressed variant (Nemotron-Parse-1.1-TC) differ from the full model?\",\"answer\":\"Nemotron-Parse-1.1-TC uses a reduced vision token length to reduce latency and computation. It offers about a 20% speed improvement with minimal quality degradation and maintains core document understanding features.\"}]","Nemotron-Parse 1.1 - Lightweight Document Parsing and OCR Model | PDF",1792981483,35]