Datalab 发布了 Marker 2，这是一个经过重写的大幅改进的开源文档转换管道。新版本与 MinerU 相比，吞吐量提高了 5 倍，在单个 Nvidia B200 GPU 上达到每秒 2.9 页，并在 olmOCR-bench 上获得 76.0% 的分数。Marker 2 使用 Surya OCR 2 和更快的 pdftext 模块这三个核心组件重建，并提供针对质量、速度或纯 CPU 操作进行了优化的多种转换路径。 AI
影响 此版本在文档处理速度和准确性方面提供了显著改进，可能影响依赖于将各种文件格式转换为结构化数据的流程。
排序理由 发布了一个新的开源文档转换管道版本。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
<p>Datalab rewrote Marker as a three-mode pipeline. Version 2 hits 76.0 on olmOCR-bench and sustains 2.9 pages per second on one B200 — over 5× MinerU's pipeline backend, while beating Docling on both accuracy and speed. Here's how it compares against MinerU, Docling and LitePars…
<p>Datalab rewrote Marker as a three-mode pipeline. Version 2 hits 76.0 on olmOCR-bench and sustains 2.9 pages per second on one B200 — over 5× MinerU's pipeline backend, while beating Docling on both accuracy and speed. Here's how it compares against MinerU, Docling and LitePars…
<p>Datalab rewrote Marker as a three-mode pipeline. Version 2 hits 76.0 on olmOCR-bench and sustains 2.9 pages per second on one B200 — over 5× MinerU's pipeline backend, while beating Docling on both accuracy and speed. Here's how it compares against MinerU, Docling and LitePars…