Datalab Marker v2 vs MinerU, Docling, and Liteparse: Benchmark Breakdown

Datalab released Marker 2, a full rewrite of its open-source document conversion system.

The new version adds three conversion modes, better OCR and layout handling, CPU support, and a shared inference-server design.

On olmOCR-bench, balanced mode scored 76.0% and processed 2.9 pages per second on a single B200 GPU, outperforming MinerU and Docling in the reported comparison.

The release also introduces breaking packaging and Python-version changes, with some structured-extraction features removed, so teams should validate it on their own documents.