Tasks and inputs
WeVisDoc 4B is listed for Image-to-text, Document OCR. It accepts image, text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
WeVisDoc 4B converts document-page images into structured text, including formulas and tables.
The specification score summarizes recorded features and limits. It is not a benchmark result.
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
WeVisDoc 4B converts document-page images into structured text, including formulas and tables.
Tencent positions this 4B model for a larger page parser to compare on complex layouts. Its documented outputs include Markdown, LaTeX formulas and HTML tables. The release uses Apache 2.0 and is based on a vision-language backbone. It is a page-understanding model rather than a general hosted OCR service with an established per-page price.
Use scans containing paragraphs, multi-column text, mathematical notation and a table with merged cells. Compare reading order, headings, numeric values, row associations and formula structure with the source. Include a poor scan and a page with a dense footer. Review valid markup separately from accurate content: a syntactically correct table can still assign a value to the wrong row.
Keep page ordering and document boundaries in the surrounding application. Render PDFs with a consistent resolution and record any crop or rotation before inference. Validate output with the downstream Markdown, formula or table parser you intend to use. A runtime example’s maximum-length setting is not necessarily the model’s universal context limit, so this profile does not import that setting as an architecture fact.
For deployment, follow the publisher’s processor and prompt format and measure complete processing time per page, including rendering and validation. Compare accepted extraction quality before selecting a model based only on size.
Compare the model on the work you need to complete.
WeVisDoc 4B is listed for Image-to-text, Document OCR. It accepts image, text 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 Apache 2.0. 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.
Mistral AI
Mistral OCR 4.1 is a document-processing service for extracting text, structure and layout information from supplied pages.
deepseek-ai
Hugging Face repository for Image-text-to-text.
paddlepaddle
PaddleOCR VL from paddlepaddle: Compact vision-language model for multilingual document parsing, OCR, tables, formulas, and charts.
Tencent
WeVisDoc 2B converts document-page images into structured text, including formulas and tables.