OCRRouter

Marker vs olmOCR

Pick Marker if you want a PDF to Markdown converter that also runs on CPU or Apple Silicon and your company falls under the weight licence terms. Pick olmOCR if you have an NVIDIA GPU with 12 GB or more and want fully open weights and a low per-page cost at scale. The main split is hardware and licence: Marker runs almost anywhere but restricts commercial use above $5M, while olmOCR needs a GPU but is fully open.

Markerdatalab/markerolmOCRallenai/olmocr
TypeOCR APIOpen source
Price14 credits / page2 credits / page
Per 1,000 pages$14.00$2.00
Typical speed2.5 s / page1.5 s / page
Best forBooks and research papersBulk PDF to text for datasets
InputsPDF, images, DOCXPDF, images
Languages90+ languagesMainly English
Max pages per request1,0001,000
Tables✓ Yes✓ Yes
HandwritingNo✓ Yes
Math✓ Yes✓ Yes
Bounding boxes✓ YesNo
ConfidenceNoNo
FormsNoNo
Multilingual✓ YesNo
On OCRRouterAvailableAvailable

Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.

Which to choose

Choose Marker if

  • You have no NVIDIA GPU: Marker runs on CPU or Apple Silicon, and its fast mode runs on CPU with about 18 MB of ONNX weights at 23.6 pages per second.
  • You need more than Markdown: it outputs HTML, JSON block trees, RAG chunks, LaTeX and images.
  • You want a hosted option from the same vendor: Datalab's API includes about 5,000 free pages per month.

Choose olmOCR if

  • You need free commercial use at any company size: the toolkit is Apache-2.0 and weights, training data and code are all released.
  • You process PDFs at volume: self-hosting costs under $200 per million pages, per Ai2.
  • Your pages mix printed text with handwriting, tables and equations, and you want headers and footers stripped in reading order.

Key differences

MarkerolmOCR
LicenceCode Apache-2.0; weights under a modified AI Pubs Open Rail-M licence, free for research, personal use and startups under $5MApache-2.0 toolkit; weights, training data and code released
HardwareGPU, CPU or MPS; vllm mode needs an NVIDIA GPUNVIDIA GPU with at least 12 GB VRAM and 30 GB free disk; no CPU-only use
PriceFree self-hosted within the weight licence; hosted API with about 5,000 free pages per monthUnder $200 per million pages self-hosted; DeepInfra hosted at $0.09 input / $0.19 output per million tokens
Output formatsMarkdown, HTML, JSON, chunks, LaTeX, imagesMarkdown, JSON
olmOCR-Bench score83.7 in vllm mode (Datalab's own number); 66.6 in CPU fast mode82.4 (v0.4.0), listed above Marker at 76.1 in the same table
HandwritingNot publishedHandled, along with equations and tables
Known failureCPU fast mode scores lower; weights block free commercial use for companies above $5MOut of memory on a 12 GB GPU (issue #424), stuck on the last few pages (issue #433), unreliable image bounding boxes (issue #437)

Sources

What each one is

Marker

Hosted version of the open-source Marker and Surya models. Strong on books and papers.

Marker details

olmOCR

Open-source vision model from Ai2 (often searched as olm ocr) that turns PDFs into clean, reading-order text. Latest release: olmOCR 2.

olmOCR details

Try both on your own file

Run the file once with each model and compare the output.

PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?

Use one, fall back to the other

{"model": "datalab/marker","fallbacks": ["allenai/olmocr"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

Marker alternativesolmOCR alternativesBest OCR models in 2026