Marker 2.0.0

Marker 2.0.0

Marker 2 is a rewrite focused on speed, full CPU support, and accuracy. It's built on three new pieces we shipped over the last few months: Surya OCR 2, a 20M-param fast layout model, and a rebuilt (3× faster) pdftext.

Against comparable pipeline OCR systems, Marker 2 is both faster and more accurate — on olmOCR-bench, balanced mode scores 76.0% overall (83.5% on born-digital) while running over 5× more pages/sec than MinerU's pipeline backend.

Highlights

• Conversion modes. Choose your speed/accuracy tradeoff:
  • balanced — surya VLM for layout + full-page OCR when text is bad. Highest quality; best on GPU. 76.0% olmOCR-bench.
  • fast — lightweight rf-detr/onnx layout + pdftext, with minimal, surgical VLM use. 66.6%, and far cheaper.
  • --disable_ocr — pure text-layer extraction, no VLM at all. Runs entirely on CPU. 43.6%, 23.7 pg/s.
• Device-aware default. Mode now defaults by device automatically — balanced on GPU, fast on CPU/MPS — so you get a sensible speed/accuracy tradeoff without a flag. Override any time with --mode.
• Full CPU support. fast --disable_ocr needs no GPU and no inference server; the 20M layout model runs on CPU and still reads structure (columns, tables, headers).
• Throughput by design. Many thin CPU workers share a single surya inference server; the parent process budgets VLM concurrency across them, so throughput scales with server capacity instead of per-process VRAM. Sustained on one B200: 2.9 pg/s balanced, 7.4 pg/s fast, 23.7 pg/s no-OCR.
• Selective OCR. Marker reads the PDF text layer with pdftext and only calls the VLM where it's needed — garbled/scanned pages, equations, low-confidence tables — instead of VLM-OCRing every page.
• Reproducible benchmarks. A new benchmarks/ harness reproduces every olmOCR-bench score and throughput number in the README, for marker and competitors (MinerU, docling, liteparse).

Breaking changes

• Python 3.10+ required.
• Packaging moved from Poetry to uv (build backend is now hatchling). pip install marker-pdf is unchanged.
• The unused structured-extraction converter/extractors were removed; use the Datalab API or a --use_llm workflow for structured extraction.
• Mode default is now device-dependent rather than always balanced (see above).

Install

pip install marker-pdf          # PDFs
pip install marker-pdf[full]    # + docx, pptx, xlsx, epub, html

Benchmarks (olmOCR-bench, macro-average over 8 categories)

Mode	Overall	Digital-only	Throughput (B200)
balanced (GPU)	76.0	83.5	2.9 pg/s
fast (GPU)	66.6	71.6	7.4 pg/s
fast, no OCR (CPU)	43.6	55.8	23.7 pg/s

Full per-category scores, the competitive comparison, and reproduction steps are in the README.

For the highest possible accuracy, the hosted Datalab API runs Chandra (our document VLM) with automatic correction - free $5 in credits to try.

v1.10.2

• Bump surya for upstream fix - https://github.com/datalab-to/surya/releases/tag/v0.17.1

What's Changed

• update readme by @u-ashish in #908
• Dev by @u-ashish in #909
• Fix license by @VikParuchuri in #926
• Add model license by @VikParuchuri in #928
• Tarun/surya latexfix bump by @tarun-menta in #980
• Dev by @tarun-menta in #981

Full Changelog: v1.10.1...v1.10.2

Minor fixes

What's Changed

• Fix typo in superscript/subscript condition check by @Tenkeboks in #897
• Enable passing arbitrary config by @VikParuchuri in #901
• Dev by @VikParuchuri in #902

New Contributors

• @Tenkeboks made their first contribution in #897

Full Changelog: v1.10.0...v1.10.1

New Layout Model + Misc Updates

Model Update

• Upgrade to a new layout model through surya. Major performance boost

Misc Updates

• README updates
• Added a new flag - --html_tables_in_markdown. When output_format is set to markdown, this will render tables using html tags, instead of the default markdown syntax.

What's Changed

• small copy + readme update by @u-ashish in #879
• Dev by @u-ashish in #880
• Layout release by @tarun-menta in #892
• Dev by @tarun-menta in #893

Full Changelog: v1.9.3...v1.10.0

Enable metadata storage

What's Changed

• Update README by @u-ashish in #872
• Add Modal example for marker deployment by @u-ashish in #850
• Dev by @u-ashish in #873
• Vik/quality by @VikParuchuri in #876
• Dev by @VikParuchuri in #877

New Contributors

• @u-ashish made their first contribution in #872

Full Changelog: v1.9.2...v1.9.3

v1.9.2

Misc Updates

• Allow LLM processor to loop to improve tables further
• Detect and fix cases where table cells cut text lines to reduce hallucinations
• Update commercial terms

What's Changed

• Table Hotfixes by @tarun-menta in #865
• Vik/table loop by @VikParuchuri in #864
• Updated commercial language by @sandy0kwon in #843
• Add image format to img_to_base64 in BaseService by @EdmondChuiHW in #869
• Dev by @VikParuchuri in #868

New Contributors

• @EdmondChuiHW made their first contribution in #869

Full Changelog: v1.9.1...v1.9.2

Fix Blank Table Cells

What's Changed

• fix: make sure rounded poly == blank if all same coords by @zanussbaum in #857
• fix: blank table cells by @zanussbaum in #861

Full Changelog: v1.9.0...v1.9.1

Marker Block Mode

Moving marker to block mode inference. OCR is done at the block level now, instead of the line level. While this is a bit slower, it boosts accuracy.

What's Changed

• Marker Block Mode by @tarun-menta in #831
• Dev by @tarun-menta in #856

Full Changelog: v1.8.5...v1.9.0

Gemini JSON fix

What's Changed

• update license and README to reflect OpenRAIL license change by @sandy0kwon in #844
• Dev by @VikParuchuri in #848

New Contributors

• @sandy0kwon made their first contribution in #844

Full Changelog: v1.8.4...v1.8.5

Misc fixes

What's Changed

• fix: increase max tokens for equation processor by @zanussbaum in #828
• Add block ids to html renderer by @VikParuchuri in #840
• Fix: retry on invalid JSON from Gemini by @runarmod in #829
• Add disable_ocr_math to table processor (2) by @ArnoKlein in #826
• Fix: show tqdm total iteration count by @runarmod in #798
• Optional block ids by @VikParuchuri in #842

New Contributors

• @ArnoKlein made their first contribution in #826

Full Changelog: v1.8.3...v1.8.4