opendatalab/MinerU mineru-2.6.3-released on GitHub

What's Changed

2025/10/31 2.6.3 Release
- Added support for a new backend vlm-mlx-engine, enabling MLX-accelerated inference for the MinerU2.5 model on Apple Silicon devices. Compared to the vlm-transformers backend, vlm-mlx-engine delivers a 100%–200% speed improvement.
- Bug fixes: #3849, #3859

Parsing Backend	pipeline (Accuracy¹ 82+)	vlm (Accuracy¹ 90+)
Parsing Backend	pipeline (Accuracy¹ 82+)	transformers	mlx-engine	vllm-engine / vllm-async-engine	http-client
Backend Features	Fast, no hallucinations	Good compatibility, but slower	Faster than transformers	Fast, compatible with the vLLM ecosystem	Suitable for OpenAI-compatible servers⁵
Operating System	Linux² / Windows / macOS		macOS³	Linux² / Windows⁴	Any
CPU inference support	✅		❌		Not required
GPU Requirements	Volta or later architectures, 6 GB VRAM or more, or Apple Silicon		Apple Silicon	Volta or later architectures, 8 GB VRAM or more	Not required
Memory Requirements	Minimum 16 GB, 32 GB recommended				8 GB
Disk Space Requirements	20 GB or more, SSD recommended				2 GB
Python Version	3.10-3.13

¹ Accuracy metric is the End-to-End Evaluation Overall score of OmniDocBench (v1.5), tested on the latest MinerU version.
² Linux supports only distributions released in 2019 or later.
³ MLX requires macOS 13.5 or later, recommended for use with version 14.0 or higher.
⁴ Windows vLLM support via WSL2(Windows Subsystem for Linux).
⁵ Servers compatible with the OpenAI API, such as local or remote model services deployed via inference frameworks like vLLM, SGLang, or LMDeploy.

2025/10/31 2.6.3 发布
- 增加新后端vlm-mlx-engine支持，在Apple Silicon设备上支持使用MLX加速MinerU2.5模型推理，相比vlm-transformers后端，vlm-mlx-engine后端速度提升100%~200%。
- bug修复: #3849 #3859

解析后端	pipeline (精度¹ 82+)	vlm (精度¹ 90+)
解析后端	pipeline (精度¹ 82+)	transformers	mlx-engine	vllm-engine / vllm-async-engine	http-client
后端特性	速度快, 无幻觉	兼容性好, 速度较慢	比transformers快	速度快, 兼容vllm生态	适用于OpenAI兼容服务器⁵
操作系统	Linux² / Windows / macOS		macOS³	Linux² / Windows⁴	不限
CPU推理支持	✅		❌		不需要
GPU要求	Volta及以后架构, 6G显存以上或Apple Silicon		Apple Silicon	Volta及以后架构, 8G显存以上	不需要
内存要求	最低16GB以上, 推荐32GB以上				8GB
磁盘空间要求	20GB以上, 推荐使用SSD				2GB
python版本	3.10-3.13

¹ 精度指标为OmniDocBench (v1.5)的End-to-End Evaluation Overall分数，基于MinerU最新版本测试
² Linux仅支持2019年及以后发行版
³ MLX需macOS 13.5及以上版本支持，推荐14.0以上版本使用
⁴ Windows vLLM通过WSL2(适用于 Linux 的 Windows 子系统)实现支持
⁵ 兼容OpenAI API的服务器，如通过vLLM/SGLang/LMDeploy等推理框架部署的本地模型服务器或远程模型服务

New Contributors

@xvlincaigou made their first contribution in #3841
@aopstudio made their first contribution in #3870
@baymax2099 made their first contribution in #3880

Full Changelog: mineru-2.6.2-released...mineru-2.6.3-released