---

license: mit

language:

• en
• zh
• multilingual

base_model:

• dots-studio/dots.mocr

library_name: transformers

tags:

• text-generation-inference
• llama-cpp
• image-to-text
• ocr
• document-parse
• layout
• table
• formula
• transformers
• custom_code

pipeline_tag: image-text-to-text

---

dots.mocr-GGUF

> dots.mocr is an advanced multimodal OCR model developed by rednote-hilab as the successor to dots.ocr, built on a 3B-parameter vision-language model foundation that extends beyond standard document parsing to unify layout detection, content recognition, structured graphics parsing, grounding, semantic understanding, and interactive dialogue within a single framework. It achieves state-of-the-art performance among models of comparable size across multiple benchmarks, including OmniDocBench (v1.5), olmOCR-bench (83.9%), and XDocParse, surpassing competing specialized models like MonkeyOCR-pro-3B, GLM-OCR, PaddleOCR-VL-1.5, and HuanyuanOCR, while approaching the performance of much larger general VLMs like Gemini 3 Pro. A distinctive capability of dots.mocr is its ability to parse structured graphics — including charts, UI layouts, scientific figures, chemical formulas, and logos — directly into SVG code, with a companion model dots.mocr-svg specifically optimized for this image-to-SVG task, achieving scores of 0.902 on UniSVG, 0.905 on ChartMimic, and 0.901 on ChemDraw. The model supports multilingual document parsing across 100+ languages, handles diverse document types, outputs structured JSON with bounding boxes and layout categories, and supports inference via both HuggingFace Transformers and vLLM (officially integrated since vLLM v0.11.0), with additional capabilities including web parsing, scene text spotting, and general visual question answering.

Model Files

File Name | Quant Type | File Size | File Link |

|-----------|------------|-----------|-----------|

| dots.mocr.BF16.gguf | BF16 | 3.56 GB | Download |

| dots.mocr.F16.gguf | F16 | 3.56 GB | Download |

| dots.mocr.Q3_K_L.gguf | Q3_K_L | 980 MB | Download |

| dots.mocr.Q3_K_M.gguf | Q3_K_M | 924 MB | Download |

| dots.mocr.Q3_K_S.gguf | Q3_K_S | 861 MB | Download |

| dots.mocr.Q4_K_M.gguf | Q4_K_M | 1.12 GB | Download |

| dots.mocr.Q4_K_S.gguf | Q4_K_S | 1.07 GB | Download |

| dots.mocr.Q5_K_M.gguf | Q5_K_M | 1.29 GB | Download |

| dots.mocr.Q5_K_S.gguf | Q5_K_S | 1.26 GB | Download |

| dots.mocr.Q6_K.gguf | Q6_K | 1.46 GB | Download |

| dots.mocr.Q8_0.gguf | Q8_0 | 1.89 GB | Download |

| dots.mocr.mmproj-bf16.gguf | mmproj-bf16 | 2.53 GB | Download |

| dots.mocr.mmproj-f16.gguf | mmproj-f16 | 2.53 GB | Download |

| dots.mocr.mmproj-q8_0.gguf | mmproj-q8_0 | 1.34 GB | Download |

llama.cpp

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