Tasks and inputs
Youtu-Parsing-Omni is listed for Image-to-text, Multimodal, Document OCR. It accepts image, audio, video and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Youtu-Parsing-Omni is Tencent Youtu's 5B open parser that turns documents, charts, images, audio and video into one structured JSON format.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Pricing conditions: No API pricing stated; open weights for self-hosting
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
Ingestion pipelines rarely have one input type. A knowledge base gets PDFs, slide screenshots, flowcharts, meeting recordings and screen captures, each traditionally handled by a different parser with its own output format. Youtu-Parsing-Omni handles all of them with one 5B model and one JSON envelope.
Tencent Youtu released the weights, technical report, a vLLM plugin and examples in October 2026.
| Task flag | Input | Main contents of the output |
|---|---|---|
document |
Document page | Layout elements with boxes, text, LaTeX, tables, charts as Markdown, flowcharts as Mermaid, reading order |
natural_image |
Photo | Entities and text with boxes, tags, captions |
graphics_chart |
Chart | Markdown table, notes, caption |
graphics_flowchart |
Flowchart | Mermaid diagram plus caption |
graphics_geometric |
Geometry figure | Points, lines, arcs, shapes, relations, measurements |
audio |
Audio clip | Timestamped segments, speakers, ASR, acoustic events |
natural_video |
Video | Segments, actions, camera motion, audio track |
textrich_video |
Screen or slide video | OCR plus ASR and a Markdown report |
The envelope covers perception (what is where) and cognition (captions, narratives, reports). The task prompt picks the family.
Tencent reports 96.96 overall on OmniDocBench v1.6, which it describes as state of the art, and an average of 75.08 on OmniParsingBench, the best among open-weight models and second only to Gemini 3 Pro in its table. Those are publisher numbers.
Not Apache. The model uses a custom Youtu-Parsing licence. Read it before building a commercial product, since parsing tends to sit deep inside paid pipelines.
Build a test folder that matches your intake: thirty scanned pages, ten charts, five flowcharts, a few audio files and two screen recordings. Validate that every output parses as JSON and matches your downstream schema, then score table accuracy and reading order on the documents. One model replacing four is only a win if the weakest modality is still good enough. Compare document results against WeVisDoc 2B and the rest of the document OCR directory.
Compare the model on the work you need to complete.
Youtu-Parsing-Omni is listed for Image-to-text, Multimodal, Document OCR. It accepts image, audio, video and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.
The weights are downloadable under Youtu-Parsing licence (other). Check runtime support and size the complete model and working context for your expected concurrency.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
OpenBMB
MiniCPM-OCR 1.0 is OpenBMB's open 1.3B document-parsing model for text, tables, formulas and charts, with a free API and on-device apps.
OpenBMB
MiniCPM-OCR Jev is OpenBMB's open 2B judge that scores whether an OCR result for a text, formula or table region is correct, from 0 to 1.
lightonai
LightOnOCR-3-0.8B is LightOn's smallest open document model: page images in, markdown out, with optional bounding boxes and chart tables.
Sarvam
Sarvam Vision 2.1 is Sarvam's 3B document-intelligence VLM for OCR, tables, forms and Indic handwriting in 22 Indian languages plus English.
NVIDIA
Nemotron 3.5 Parse is NVIDIA's 0.631B open document parser that turns page images into Markdown, HTML or JSON with tables, charts and reading order.
Cohere
Cohere Parse (parse-v5.0) is a 2.3B vision-language document parser that turns PDFs, slides and scans into Markdown for RAG, at $1.50 per 1,000 pages.