---

base_model:

• tencent/WeVisDoc-4B

tags:

• text-generation-inference
• llama-cpp
• qwen3_vl

license: apache-2.0

language:

• en
• zh

pipeline_tag: image-text-to-text

library_name: transformers

---

WeVisDoc-4B-GGUF

> WeVisDoc-4B is an end-to-end document parsing model developed by Tencent that converts page images directly into structured Markdown output, complete with LaTeX-formatted formulas and HTML tables. Fine-tuned from Qwen3-VL-4B-Instruct, it is designed specifically for document parsing (rather than being a general-purpose vision-language model) and supports both English and Chinese. The model achieves state-of-the-art results among end-to-end document parsing specialists, scoring 95.38 Overall on OmniDocBench v1.6 and a mean Overall score of 75.54 across the three PureDocBench tracks (Clean, Digital Degraded, and Real Degraded), ranking first in all four reported settings while outperforming larger models like olmOCR-2-7B. Released under the Apache 2.0 license, it can be deployed via vLLM (requiring vLLM ≥0.11.1) for efficient serving or run locally with Transformers, and is accompanied by a smaller 2B variant (WeVisDoc-2B) for lighter deployments.

Model Files

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

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

| WeVisDoc-4B.BF16.gguf | BF16 | 8.83 GB | Download |

| WeVisDoc-4B.F16.gguf | F16 | 8.83 GB | Download |

| WeVisDoc-4B.Q3_K_L.gguf | Q3_K_L | 2.41 GB | Download |

| WeVisDoc-4B.Q3_K_M.gguf | Q3_K_M | 2.24 GB | Download |

| WeVisDoc-4B.Q4_K_M.gguf | Q4_K_M | 2.72 GB | Download |

| WeVisDoc-4B.Q4_K_S.gguf | Q4_K_S | 2.6 GB | Download |

| WeVisDoc-4B.Q5_K_M.gguf | Q5_K_M | 3.16 GB | Download |

| WeVisDoc-4B.Q5_K_S.gguf | Q5_K_S | 3.09 GB | Download |

| WeVisDoc-4B.Q6_K.gguf | Q6_K | 3.63 GB | Download |

| WeVisDoc-4B.Q8_0.gguf | Q8_0 | 4.69 GB | Download |

| WeVisDoc-4B.mmproj-bf16.gguf | mmproj-bf16 | 839 MB | Download |

| WeVisDoc-4B.mmproj-f16.gguf | mmproj-f16 | 839 MB | Download |

| WeVisDoc-4B.mmproj-q8_0.gguf | mmproj-q8_0 | 454 MB | Download |

llama.cpp

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