---
title: PP-OCRv6-small-rec-GGUF
canonical_url: "https://www.modelscope.cn/models/mlx-community/PP-OCRv6-small-rec-GGUF"
md_url: "https://www.modelscope.cn/models/mlx-community/PP-OCRv6-small-rec-GGUF.md"
repository: mlx-community/PP-OCRv6-small-rec-GGUF
last_updated: 2026-07-31
license: apache-2.0
library_name:
  - gguf
  - pytorch
frameworks:
  - pytorch
downloads: 18
stars: 0
tags:
  - ocr
  - pp-ocrv6
  - gguf
  - crispembed
---

# PP-OCRv6-small-rec-GGUF

> PP-OCRv6-small-rec-GGUF - mlx-community 在 ModelScope 开源的模型。PP-OCRv6 small recognizer — CrispEmbed GGUF

mlx-community/PP-OCRv6-small-rec-GGUF 是 ModelScope 魔搭社区上的机器学习模型，采用 apache-2.0 许可。

- **Repository**: mlx-community/PP-OCRv6-small-rec-GGUF
- **License**: apache-2.0
- **Tags**: ocr, pp-ocrv6, gguf, crispembed
- **Downloads**: 18
- **Stars**: 0
- **Last updated**: 2026-07-31

Source: https://www.modelscope.cn/models/mlx-community/PP-OCRv6-small-rec-GGUF

---

# PP-OCRv6 small recognizer — CrispEmbed GGUF

Files: `PP-OCRv6_small_rec-f16.gguf` and `PP-OCRv6_small_rec-crispasr-q4_k-policy.gguf`.

The policy-q4 container intentionally keeps the complete PP-OCRv6 recognizer graph in F16: quantizing intermediate CNN/SVTR weights caused compounding CTC drift. Source: PaddlePaddle PP-OCRv6, Apache-2.0.

Parity on `tests/regression/images/fox.png`: F16 and policy-q4 both pass the 0.999 cosine floor across the full native diff path; worst stage cosine is 0.999265 and logits cosine is 0.999530.