---

license: apache-2.0

base_model:

- Logics-MLLM/Logics-Parsing-v2

base_model_relation: quantized

pipeline_tag: image-text-to-text

library_name: gguf

language:

- zh

- en

tags:

- ocr

- document-parsing

- gguf

- llama.cpp

- 42model

---

<!-- README for the Hugging Face repo 42ailab/Logics-Parsing-v2-GGUF. Editing this file changes the repo landing page. -->

Logics-Parsing-v2 · A 4B Document Parser for Formulas, Tables and Flowcharts

English | 简体中文 · Also on ModelScope · 42ailab.com

Give it a page (a book, a paper, a report) and it returns structured HTML: headings, paragraphs, tables and formulas marked separately, flowcharts as Mermaid, sheet music as ABC notation. It runs entirely on your own machine.

> The model was built by the Alibaba Logics team and released under Apache-2.0 (Logics-Parsing-v2, based on Qwen3-VL-4B). This repository is neither a new model nor our conversion. The GGUF files were made by community packager mradermacher; this repository is a byte-for-byte backup of them.

Why a backup

The original GGUF lives under a single personal account, on a single host. If it is renamed, deleted or re-uploaded, people who depend on it lose their download, and access to Hugging Face from mainland China is unreliable.

So we keep the original files, unmodified, on ModelScope and on our own Hugging Face account. File names, sizes and sha256 digests all match the original.

What we checked

Byte-identical: every file's sha256 matches mradermacher's repository.

Reproducible: we also converted and quantized the official Alibaba weights ourselves. The quantized data is not byte-identical (different quantizer versions), but on our two test pages (a Chinese book page and an English paper page) both builds produce identical output, layout coordinates included. That confirms the original conversion is sound, and that we could rebuild an equivalent file from the official weights if the original ever disappeared.

This comparison covers two pages. It shows the conversion is correct; it is not an evaluation of the model. For model quality, see the upstream benchmarks (LogicsDocBench, OmniDocBench v1.5).

Use

Packaged for 42model: download Logics-Parsing-v2 from the OCR section of the desktop app, or run 42model download logics-parsing-v2.

Files and license

| File | Role | Size |

|---|---|---:|

| Logics-Parsing-v2.Q4_K_M.gguf | text decoder (default) | 2.33 GiB |

| Logics-Parsing-v2.Q5_K_M.gguf | text decoder | 2.69 GiB |

| Logics-Parsing-v2.Q6_K.gguf | text decoder | 3.08 GiB |

| Logics-Parsing-v2.Q8_0.gguf | text decoder | 3.99 GiB |

| Logics-Parsing-v2.mmproj-f16.gguf | vision encoder (shared) | 0.78 GiB |

Weights are under the upstream Apache-2.0 license. Credit for the GGUF conversion goes to mradermacher.