OCRRouter

Marker vs EasyOCR

Pick Marker to turn PDFs such as books and papers with equations into Markdown, HTML or JSON, and to chunk them for RAG. Pick EasyOCR for short text in photos or quick multilingual prototypes that need boxes and confidence scores. The main split is scope: Marker handles full documents, while EasyOCR returns plain text and boxes and is not meant for page layout or tables.

Markerdatalab/markerEasyOCRjaided/easyocr
TypeOCR APIOpen source
Price14 credits / page1 credit / page
Per 1,000 pages$14.00$1.00
Typical speed2.5 s / page0.9 s / page
Best forBooks and research papersQuick scripts and scene text
InputsPDF, images, DOCXimages
Languages90+ languages80+ languages
Max pages per request1,000100
Tables✓ YesNo
HandwritingNoNo
Math✓ YesNo
Bounding boxes✓ Yes✓ Yes
ConfidenceNo✓ Yes
FormsNoNo
Multilingual✓ Yes✓ Yes
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 need PDF to Markdown, HTML, JSON block trees or RAG chunks, not just raw text.
  • Your documents are books or papers with equations, and Datalab reports 83.7 on olmOCR-bench in vllm mode.
  • You have no GPU and can use fast mode, which runs on CPU at 23.6 pages per second, though its olmOCR-bench score drops to 66.6.

Choose EasyOCR if

  • You are reading scene text in photos or short snippets and want bounding boxes with confidence scores.
  • You need one of 80+ languages, including Latin, Chinese, Arabic, Devanagari and Cyrillic scripts, in a quick prototype.
  • You want plain Apache-2.0 licensing with no revenue cap and a three-line Python setup that works on CPU with gpu=False.

Key differences

MarkerEasyOCR
What it isPDF to Markdown converter from Datalab, built on the Surya OCR models. Datalab also sells a hosted API.Python OCR library from Jaided AI using CRAFT detection and a CRNN recognizer. Library only, no hosted option.
LicenceCode is Apache-2.0. Model weights use a modified AI Pubs Open Rail-M license, free for research, personal use and startups under $5M. Larger companies need a paid license for commercial use.Apache-2.0.
Output formatsMarkdown, HTML, JSON, chunks, LaTeX and images.Plain text and bounding boxes.
HardwareGPU, CPU or Apple Silicon (MPS). vllm mode needs an NVIDIA GPU. Fast mode runs on CPU with about 18 MB of ONNX weights.PyTorch on CPU (gpu=False) or NVIDIA GPU. The default reader fails on Apple M1 unless GPU is disabled (issue #1136).
PriceFree self-hosted within the weight license, you pay for compute. Hosted API includes about 5,000 free pages per month, with paid plans on Datalab's pricing page.Free self-hosted, you pay for compute.
BenchmarksolmOCR-bench 83.7 and OmniParseBench 85.9 in vllm mode. All figures are Datalab's own.Not published
Known failure modesCPU fast mode scores 66.6 on olmOCR-bench, well below vllm mode.No text detected after upgrading to 1.7.0 (issue #1039). Gibberish when batch_size is above 1 (issue #1038). Trouble reading the digit 1 (issue #747). Issues older than 6 months are auto-closed.

Sources

What each one is

Marker

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

Marker details

EasyOCR

Python OCR library with a simple API. Returns text lines with boxes and confidence.

EasyOCR 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": ["jaided/easyocr"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

Marker alternativesEasyOCR alternativesBest OCR models in 2026