dots.ocr

[![HuggingFace](https://img.shields.io/badge/HuggingFace%20Weights-black.svg?logo=HuggingFace)](https://huggingface.co/rednote-hilab/dots.ocr-1.5) [![Arxiv](https://img.shields.io/badge/arXiv-Paper-B31B1B.svg?logo=arxiv)](https://arxiv.org/abs/2512.02498)

🖥️ Live Demo | 💬 WeChat | 📕 rednote | 🐦 X

## Introduction **dots.ocr** Designed for universal accessibility, it possesses the capability to recognize virtually any human script. Beyond achieving state-of-the-art (SOTA) performance in standard multilingual document parsing among models of comparable size, dots.ocr-1.5 excels at converting structured graphics (e.g., charts and diagrams) directly into SVG code, parsing web screens and spotting scene text. ## News * ```2026.2.16 ``` 🚀 We release [dots.ocr-1.5](https://huggingface.co/rednote-hilab/dots.ocr-1.5), designed to recognize all human scripts and symbols. This model extends beyond standard document parsing to include comprehensive image parsing. We are simultaneously releasing [dots.ocr-1.5-svg](https://huggingface.co/rednote-hilab/dots.ocr-1.5-svg), which offers more robust performance for image parsing tasks. * ```2025.10.31 ``` 🚀 We release [dots.ocr.base](https://huggingface.co/rednote-hilab/dots.ocr.base), foundation VLM focus on OCR tasks, also the base model of [dots.ocr](https://huggingface.co/rednote-hilab/dots.ocr). Try it out! * ```2025.07.30 ``` 🚀 We release [dots.ocr](https://huggingface.co/rednote-hilab/dots.ocr), — a multilingual documents parsing model based on 1.7b llm, with SOTA performance. ## Evaluation ### 1. Document Parsing #### 1.1 Elo Score of different bench between latest models

models	olmOCR-Bench	OmniDocBench (v1.5)	XDocParse
GLM-OCR	859.9	937.5	742.1
PaddleOCR-VL-1.5	873.6	965.6	797.6
HuanyuanOCR	978.9	974.4	895.9
dots.ocr	1027.4	994.7	1133.4
dots.ocr-1.5	1089.0	1025.8	1157.1
Gemini 3 Pro	1171.2	1102.1	1273.9

> **Notes:** > - Results for Gemini 3 Pro, PaddleOCR-VL-1.5, and GLM-OCR were obtained via APIs, while HuanyuanOCR results were generated using local inference. > - The Elo score evaluation was conducted using Gemini 3 Flash. The prompt can be found at: [Elo Score Prompt](https://github.com/rednote-hilab/dots.ocr/blob/master/tools/elo_score_prompt.py). These results are consistent with the findings on [ocrarena](https://www.ocrarena.ai/battle). #### 1.2 olmOCR-bench

Model	ArXiv	Old scans math	Tables	Old scans	Headers & footers	Multi column	Long tiny text	Base	Overall
Mistral OCR API	77.2	67.5	60.6	29.3	93.6	71.3	77.1	99.4	72.0±1.1
Marker 1.10.1	83.8	66.8	72.9	33.5	86.6	80.0	85.7	99.3	76.1±1.1
MinerU 2.5.4*	76.6	54.6	84.9	33.7	96.6	78.2	83.5	93.7	75.2±1.1
DeepSeek-OCR	77.2	73.6	80.2	33.3	96.1	66.4	79.4	99.8	75.7±1.0
Nanonets-OCR2-3B	75.4	46.1	86.8	40.9	32.1	81.9	93.0	99.6	69.5±1.1
PaddleOCR-VL*	85.7	71.0	84.1	37.8	97.0	79.9	85.7	98.5	80.0±1.0
Infinity-Parser 7B*	84.4	83.8	85.0	47.9	88.7	84.2	86.4	99.8	82.5±?
olmOCR v0.4.0	83.0	82.3	84.9	47.7	96.1	83.7	81.9	99.7	82.4±1.1
Chandra OCR 0.1.0*	82.2	80.3	88.0	50.4	90.8	81.2	92.3	99.9	83.1±0.9
dots.ocr	82.1	64.2	88.3	40.9	94.1	82.4	81.2	99.5	79.1% ± 1.0%
dots.ocr-1.5	85.9	85.5	90.7	48.2	94.0	85.3	81.6	99.7	83.9% ± 0.9

> **Note:** > - The metrics are from [olmocr](https://github.com/allenai/olmocr), and our own internal evaluations. > - We delete the Page-header and Page-footer cells in the result markdown. #### 1.3 Other Benchmarks

Model Type	Methods	Size	OmniDocBench(v1.5) TextEdit↓	OmniDocBench(v1.5) Read OrderEdit↓	pdf-parse-bench
GeneralVLMs	Gemini-2.5 Pro	-	0.075	0.097	9.06
	Qwen3-VL-235B-A22B-Instruct	235B	0.069	0.068	9.71
	gemini3pro	-	0.066	0.079	9.68
SpecializedVLMs	Mistral OCR	-	0.164	0.144	8.84
	Deepseek-OCR	3B	0.073	0.086	8.26
	MonkeyOCR-3B	3B	0.075	0.129	9.27
	OCRVerse	4B	0.058	0.071	--
	MonkeyOCR-pro-3B	3B	0.075	0.128	-
	MinerU2.5	1.2B	0.047	0.044	-
	PaddleOCR-VL	0.9B	0.035	0.043	9.51
	HunyuanOCR	0.9B	0.042	-	-
	PaddleOCR-VL1.5	0.9B	0.035	0.042	-
	GLMOCR	0.9B	0.04	0.043	-
	dots.ocr	3B	0.048	0.053	9.29
	dots.ocr-1.5	3B	0.031	0.029	9.54

> **Note:** > - Metrics are sourced from [OmniDocBench](https://github.com/opendatalab/OmniDocBench) and other model publications. [pdf-parse-bench](https://github.com/phorn1/pdf-parse-bench) results are reproduced by Qwen3-VL-235B-A22B-Instruct. > - Formula and Table metrics for OmniDocBench1.5 are omitted due to their high sensitivity to detection and matching protocols. ### 2. Vision-Language Parsing Visual languages (e.g., charts, graphics, chemical formulas, logos) encapsulate dense human knowledge. **dots.ocr-1.5** unifies the interpretation of these elements by parsing them directly into **SVG code**.

Methods	Unisvg			Chartmimic	Design2Code	Genexam	SciGen	ChemDraw
Methods	Low-Level	High-Level	Score	Chartmimic	Design2Code	Genexam	SciGen	ChemDraw
OCRVerse	0.632	0.852	0.763	0.799	-	-	-	0.881
Gemini 3 Pro	0.563	0.850	0.735	0.788	0.760	0.756	0.783	0.839
dots.ocr-1.5	0.850	0.923	0.894	0.772	0.801	0.664	0.660	0.790
dots.ocr-1.5-svg	0.860	0.931	0.902	0.905	0.834	0.8	0.797	0.901

> **Note:** > - We use the ISVGEN metric from [UniSVG](https://ryanlijinke.github.io/) to evaluate the parsing result. For benchmarks that do not natively support image parsing, we use the original images as input, and calculate the ISVGEN score between the rendered output and the original image. > - [OCRVerse](https://github.com/DocTron-hub/OCRVerse) results are derived from various code formats (e.g., SVG, Python), whereas results for Gemini 3 Pro and dots.ocr-1.5 are based specifically on SVG code. > - Due to the capacity constraints of a 3B-parameter VLM, dots.ocr-1.5 may not excel in all tasks yet like svg. To complement this, we are simultaneously releasing dots.ocr-1.5-svg. We plan to further address these limitations in future updates. ### 3. General Vision Tasks

Model	CharXiv_descriptive	CharXiv_reasoning	OCR_Reasoning	infovqa	docvqa	ChartQA	OCRBench	AI2D	CountBenchQA	refcoco
Qwen3vl-2b-instruct	62.3	26.8	-	72.4	93.3	-	85.8	76.9	88.4	-
dots.ocr-1.5	77.4	55.3	22.85	73.76	91.85	83.2	86.0	82.16	94.46	80.03

# Quick Start ## 1. Installation ### Install dots.ocr-1.5 ```shell conda create -n dots_ocr python=3.12 conda activate dots_ocr git clone https://github.com/rednote-hilab/dots.ocr.git cd dots.ocr # Install pytorch, see https://pytorch.org/get-started/previous-versions/ for your cuda version pip install torch==2.7.0 torchvision==0.22.0 torchaudio==2.7.0 --index-url https://download.pytorch.org/whl/cu128 pip install -e . ``` If you have trouble with the installation, try our [Docker Image](https://hub.docker.com/r/rednotehilab/dots.ocr) for an easier setup, and follow these steps: ```shell git clone https://github.com/rednote-hilab/dots.ocr.git cd dots.ocr pip install -e . ``` ### Download Model Weights > 💡**Note:** Please use a directory name without periods (e.g., `DotsOCR_1_5` instead of `dots.ocr-1.5`) for the model save path. This is a temporary workaround pending our integration with Transformers. ```shell python3 tools/download_model.py # with modelscope python3 tools/download_model.py --type modelscope ``` ## 2. Deployment ### vLLM inference We highly recommend using vLLM for deployment and inference. All of our evaluations results are based on vLLM 0.9.1 via out-of-tree model registration. **Since vLLM version 0.11.0, Dots OCR has been officially integrated into vLLM with verified performance** and you can use vLLM docker image directly (e.g, `vllm/vllm-openai:v0.11.0`) to deploy the model server. > **Note:** > - We found a little bit performance drop when using vLLM 0.11.0. We are working on a fix. ```shell # Launch vLLM model server ## dots.ocr-1.5 CUDA_VISIBLE_DEVICES=0 vllm serve rednote-hilab/dots.ocr-1.5 --tensor-parallel-size 1 --gpu-memory-utilization 0.9 --chat-template-content-format string --served-model-name model --trust-remote-code ## dots.ocr-1.5-svg CUDA_VISIBLE_DEVICES=0 vllm serve rednote-hilab/dots.ocr-1.5-svg --tensor-parallel-size 1 --gpu-memory-utilization 0.9 --chat-template-content-format string --served-model-name model --trust-remote-code # vLLM API Demo # See dots_ocr/model/inference.py for details on parameter and prompt settings # that help achieve the best output quality. ## document parsing python3 ./demo/demo_vllm.py --prompt_mode prompt_layout_all_en ## web parsing ## scene spoting ## image parsing with svg code ## general qa ``` ### Hugginface inference ```shell python3 demo/demo_hf.py ```

Hugginface inference details

```python import torch from transformers import AutoModelForCausalLM, AutoProcessor, AutoTokenizer from qwen_vl_utils import process_vision_info from dots_ocr.utils import dict_promptmode_to_prompt model_path = "./weights/DotsOCR_1_5" model = AutoModelForCausalLM.from_pretrained( model_path, attn_implementation="flash_attention_2", torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True ) processor = AutoProcessor.from_pretrained(model_path, trust_remote_code=True) image_path = "demo/demo_image1.jpg" prompt = """Please output the layout information from the PDF image, including each layout element's bbox, its category, and the corresponding text content within the bbox. 1. Bbox format: [x1, y1, x2, y2] 2. Layout Categories: The possible categories are ['Caption', 'Footnote', 'Formula', 'List-item', 'Page-footer', 'Page-header', 'Picture', 'Section-header', 'Table', 'Text', 'Title']. 3. Text Extraction & Formatting Rules: - Picture: For the 'Picture' category, the text field should be omitted. - Formula: Format its text as LaTeX. - Table: Format its text as HTML. - All Others (Text, Title, etc.): Format their text as Markdown. 4. Constraints: - The output text must be the original text from the image, with no translation. - All layout elements must be sorted according to human reading order. 5. Final Output: The entire output must be a single JSON object. """ messages = [ { "role": "user", "content": [ { "type": "image", "image": image_path }, {"type": "text", "text": prompt} ] } ] # Preparation for inference text = processor.apply_chat_template( messages, tokenize=False, add_generation_prompt=True ) image_inputs, video_inputs = process_vision_info(messages) inputs = processor( text=[text], images=image_inputs, videos=video_inputs, padding=True, return_tensors="pt", ) inputs = inputs.to("cuda") # Inference: Generation of the output generated_ids = model.generate(**inputs, max_new_tokens=24000) generated_ids_trimmed = [ out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids) ] output_text = processor.batch_decode( generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False ) print(output_text) ```

### Hugginface inference with CPU Please refer to [CPU inference](https://github.com/rednote-hilab/dots.ocr/issues/1#issuecomment-3148962536) ## 3. Document Parse **Based on vLLM server**, you can parse an image or a pdf file using the following commands: ```bash # Parse all layout info, both detection and recognition # Parse a single image python3 dots_ocr/parser.py demo/demo_image1.jpg # Parse a single PDF python3 dots_ocr/parser.py demo/demo_pdf1.pdf --num_thread 64 # try bigger num_threads for pdf with a large number of pages # Layout detection only python3 dots_ocr/parser.py demo/demo_image1.jpg --prompt prompt_layout_only_en # Parse text only, except Page-header and Page-footer python3 dots_ocr/parser.py demo/demo_image1.jpg --prompt prompt_ocr ``` **Based on Transformers**, you can parse an image or a pdf file using the same commands above, just add `--use_hf true`. > Notice: transformers is slower than vllm, if you want to use demo/* with transformers，just add `use_hf=True` in `DotsOCRParser(..,use_hf=True)`

Output Results

1. **Structured Layout Data** (`demo_image1.json`): A JSON file containing the detected layout elements, including their bounding boxes, categories, and extracted text. 2. **Processed Markdown File** (`demo_image1.md`): A Markdown file generated from the concatenated text of all detected cells. * An additional version, `demo_image1_nohf.md`, is also provided, which excludes page headers and footers for compatibility with benchmarks like Omnidocbench and olmOCR-bench. 3. **Layout Visualization** (`demo_image1.jpg`): The original image with the detected layout bounding boxes drawn on it.

## 4. Demo Have fun with the [live demo](https://dotsocr.xiaohongshu.com/). ### Examples for document parsing

### Examples for image parsing

> **Note:** > - Inferenced by dots.ocr-1.5-svg ### Example for web parsing

### Examples for scene spotting

# Limitation & Future Work - **Complex Document Elements:** - **Table&Formula**: The extraction of complex tables and mathematical formulas persists as a difficult task given the model's compact architecture. - **Picture**: We have adopted an SVG code representation for parsing structured graphics; however, the performance has yet to achieve the desired level of robustness. - **Parsing Failures:** While we have reduced the rate of parsing failures compared to the previous version, these issues may still occur occasionally. We remain committed to further resolving these edge cases in future updates. # Citation ```BibTeX @misc{li2025dotsocrmultilingualdocumentlayout, title={dots.ocr: Multilingual Document Layout Parsing in a Single Vision-Language Model}, author={Yumeng Li and Guang Yang and Hao Liu and Bowen Wang and Colin Zhang}, year={2025}, eprint={2512.02498}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2512.02498}, } ```