Tencent Releases WeVisDoc-2B and WeVisDoc-4B Document Parsing Models

Tencent Releases WeVisDoc-2B and WeVisDoc-4B Document Parsing Models

At a Glance

Item Value
Repository tencent/WeVisDoc-4B
Published 2026-09-16
License apache-2.0
Formats safetensors
Paper arXiv:2609.20423
Source type Primary source (the publisher itself)

Values determined by this site’s code at collection time. Dates are JST.

Overview

Tencent has released “WeVisDoc-2B" and “WeVisDoc-4B", end-to-end document parsing models that extract structured text from page images.

These models are fine-tuned versions based on “Qwen3-VL-2B-Instruct" and “Qwen3-VL-4B-Instruct" respectively, taking page images as input and outputting structured Markdown containing LaTeX-format mathematical formulas and HTML-format tables.

Specifications

  • Architecture: Qwen3VLForConditionalGeneration (qwen3_vl)
  • Base Models: Qwen3-VL-2B-Instruct, Qwen3-VL-4B-Instruct
  • License: apache-2.0

Performance

The measurement results of document parsing benchmarks provided by the publisher are as follows. Evaluation values represent the average of three inference runs.

OmniDocBench v1.6

→ Scroll horizontally to see all columns

Model Params Overall ↑ TextEdit ↓ FormulaCDM ↑ TableTEDS ↑ TableTEDS_S ↑ ROEdit ↓
Nanonets-OCR2* 3B 83.20 0.108 80.35 80.10 85.26 0.211
OCRFlux-3B* 3B 83.31 0.126 88.75 73.78 77.98 0.217
POINTS-Reader 3B 83.37 0.096 85.72 73.98 77.40 0.198
Nanonets-OCR-s 3B 83.61 0.108 81.46 80.18 84.51 0.213
olmOCR-2-7B* 7B 85.51 0.106 88.84 78.32 82.81 0.223
olmOCR 7B 85.74 0.139 88.10 83.00 87.17 0.216
DeepSeek-OCR* 3B 86.31 0.077 84.71 81.87 86.07 0.171
OCRVerse 4B 88.60 0.063 89.61 82.44 86.27 0.163
UniRec-0.1B* 0.1B 88.91 0.088 92.14 83.40 86.79 0.146
DeepSeek-OCR 2 3B 90.25 0.050 91.84 83.89 87.75 0.144
dots.ocr 3B 90.77 0.048 89.95 87.18 90.58 0.138
FD-RL* 4B 91.21 0.055 92.92 86.22 90.92 0.145
HunyuanOCR 1B 92.03 0.048 88.60 92.37 93.99 0.138
dots.mocr* 3B 92.57 0.042 92.09 89.78 92.92 0.133
FireRed-OCR 2B 93.26 0.037 95.44 88.04 91.06 0.131
Logics-Parsing-v2 4B 93.33 0.041 95.65 88.42 91.98 0.137
Qianfan-OCR 4B 93.90 0.040 95.08 90.53 93.31 0.130
Unlimited-OCR 3B-A0.5B 93.92 0.042 95.79 90.16 93.32 0.129
HunyuanOCR-1.5 1B 94.74 0.039 94.50 93.67 94.71 0.129
WeVisDoc-2B 2B 95.06 0.038 95.94 93.03 95.26 0.130
WeVisDoc-4B 4B 95.38 0.036 96.81 92.95 95.34 0.125

PureDocBench

→ Scroll horizontally to see all columns

Model Params Avg₃ ↑ Clean Overall ↑ Digital Degraded Overall ↑ Real Degraded Overall ↑
OCRFlux-3B 3B 42.06 47.14 41.82 37.21
DeepSeek-OCR 3B 46.98 53.50 46.95 40.48
UniRec-0.1B 0.1B 48.59 58.91 52.42 34.44
POINTS-Reader* 3B 49.24 53.78 51.24 42.69
DeepSeek-OCR-2 3B 49.51 55.53 49.41 43.60
Qianfan-OCR 4B 51.04 57.22 50.85 45.06
olmOCR-7B 7B 55.90 62.56 57.84 47.30
Nanonets-OCR2 3B 58.36 64.83 61.23 49.03
HunyuanOCR 1B 60.56 65.61 61.49 54.58
Unlimited-OCR* 3B-A0.5B 62.76 71.28 63.62 53.39
olmOCR-2-7B 7B 63.78 69.36 65.87 56.10
dots.ocr 3B 64.55 72.01 65.95 55.68
Nanonets-OCR-s* 3B 65.37 71.26 66.56 58.28
FireRed-OCR 2B 65.57 70.81 68.49 57.42
HunyuanOCR-1.5* 1B 68.79 73.98 70.81 61.59
OCRVerse 4B 69.40 73.18 71.36 63.66
dots.mocr 3B 70.39 76.27 73.16 61.73
Logics-Parsing-v2 4B 72.61 76.35 73.85 67.64
FD-RL 4B 73.92 78.38 76.33 67.04
WeVisDoc-2B 2B 73.86 79.36 76.62 65.60
WeVisDoc-4B 4B 75.54 79.81 77.74 69.08

According to the measurement results stated in the model card, WeVisDoc-4B achieves the highest scores among compared end-to-end parsers in both the OmniDocBench v1.6 Overall (95.38) and the PureDocBench 3-track average Avg₃ (75.54). It exhibits particularly high figures in OmniDocBench v1.6 FormulaCDM (96.81) and PureDocBench Clean Overall (79.81), demonstrating superior parsing accuracy for mathematical formulas and standard document images. The smaller WeVisDoc-2B is also remarkably capable, recording 95.06 in OmniDocBench v1.6 Overall and outperforming larger and other models such as HunyuanOCR-1.5 and Logics-Parsing-v2. On the other hand, in PureDocBench’s “Real Degraded Overall", WeVisDoc-2B stays at 65.60, trailing behind FD-RL (67.04) and Logics-Parsing-v2 (67.64), showing a difference in robustness toward real photographed degraded images compared to the larger 4B version.

Strengths and Use Cases

WeVisDoc specializes in generating structured Markdown from page images (PNG, JPEG, WebP) of books, research papers, and various documents. It excels not only in converting body text but also in extracting mathematical formulas using LaTeX format and analyzing table structures using HTML format. English (en) and Chinese (zh) are set as tag information, making it suitable for OCR and paper parsing use cases involving multilingual documents.

Hardware Requirements

Estimated requirements (calculated by Local Model Watch) — 4.4B parameters

Your VRAM Quantization File size Est. memory needed
12GB (RTX 4070 / 3060 12GB, etc.) BF16 9.0GB 10.8GB

Inference engine support (architecture name matched against each project’s own model registry in its source code, checked 2026-09-18): llama.cpp: registered, vLLM: registered, MLX (mlx-lm): registered.

Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference. What the quantization names mean: glossary.

How to Get It

Model files are distributed in safetensors format and can be obtained directly from Hugging Face.

An example script execution for starting a server using vLLM (>=0.11.1) is as follows:

bash scripts/serve_vllm.sh Tencent/WeVisDoc-2B

*When using the 4B model, specify Tencent/WeVisDoc-4B.

After starting the service, here is an example command to parse an image using the included client:

python -m wevisdoc.client --image page.png --output results/page.md

It can also be executed directly in a local inference environment using Transformers.

python -m wevisdoc.local --model Tencent/WeVisDoc-2B --image page.png --output results/page.md

Related Articles

Sources

New ModelsDocument Parsing,OCR,Qwen3-VL,Tencent,WeVisDoc,確認済み