prithivMLmods/dots.ocr-GGUF

dots.ocr-GGUF

dots.ocr is a multilingual document layout parsing model developed by rednote-hilab that unifies layout detection and content recognition within a single vision-language model (VLM), built upon a compact 1.7B-parameter LLM foundation (based on Qwen2.5-VL). It achieves state-of-the-art performance on OmniDocBench across text recognition, table parsing, and reading order tasks, while delivering formula recognition results comparable to much larger models like Gemini 2.5 Pro and Doubao-1.5. The model supports over 100 languages and handles diverse document types including academic papers, books, slides, financial reports, exam papers, magazines, and newspapers, outputting structured JSON with bounding boxes, layout categories (Caption, Footnote, Formula, List-item, Page-footer, Page-header, Picture, Section-header, Table, Text, Title), and extracted text formatted as LaTeX for formulas, HTML for tables, and Markdown for all other content. It can be switched between full layout parsing, detection-only, OCR-only, and grounding OCR modes simply by changing the input prompt, and supports inference via both HuggingFace Transformers and vLLM, with vLLM version 0.9.1 recommended for production deployment.

Model Files

llama.cpp

LLM inference in C/C++ — https://github.com/ggml-org/llama.cpp