Marker vs Chunkr
Pick Marker if you want PDF to Markdown on your own hardware, especially for books and papers with equations, and your company is under the $5M startup threshold or you buy a Datalab license. Pick Chunkr if you need layout analysis with bounding boxes and semantic chunks across PDF, PPT, Word and images, and you can run AGPL code or buy a commercial license. The main split is the job: Marker is a converter with published speed and accuracy numbers, Chunkr is a parsing service that also needs an LLM endpoint when self-hosted.
Markerdatalab/marker | Chunkrchunkr/chunkr | |
|---|---|---|
| Type | OCR API | OCR API |
| Price | 14 credits / page | 14 credits / page |
| Per 1,000 pages | $14.00 | $14.00 |
| Typical speed | 2.5 s / page | 3.0 s / page |
| Best for | Books and research papers | Layout-aware chunks for RAG |
| Inputs | PDF, images, DOCX | PDF, images, DOCX |
| Languages | 90+ languages | Multilingual |
| Max pages per request | 1,000 | 1,000 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | ✓ Yes | ✓ Yes |
| On OCRRouter | Available | Available |
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 want Markdown, HTML, JSON block trees, LaTeX or RAG chunks from PDFs, including books and papers with equations.
- You need to run without a GPU: fast mode runs on CPU with about 18 MB of ONNX weights at 23.6 pages per second, though it scores 66.6 on olmOCR-bench.
- You want published scores to check against: Datalab reports 83.7 on olmOCR-bench and 85.9 on OmniParseBench in vllm mode, which needs an NVIDIA GPU.
Choose Chunkr if
- You need OCR with bounding boxes and layout analysis, not just Markdown text.
- Your inputs include PPT, Word and images as well as PDF.
- You want to self-host with Docker on GPU, CPU-only or Mac ARM, and point it at any OpenAI-compatible LLM endpoint.
Key differences
| Marker | Chunkr | |
|---|---|---|
| Licence | Code is Apache-2.0. Model weights use a modified AI Pubs Open Rail-M license, free for research, personal use and startups under $5M. | AGPL-3.0, or a commercial license. Closed-source use needs the commercial license. |
| How you run it | Local on GPU, CPU or Apple Silicon. Datalab also sells a hosted API. | Cloud API, or self-host with Docker on GPU, CPU-only or Mac ARM. Self-host needs an external LLM endpoint. |
| Pricing | Free self-hosted within the weight license, you pay for compute. Hosted API includes about 5,000 free pages per month, then paid plans. | Open-source version is free, you pay for compute and an LLM endpoint. Hosted credit-based plans, official prices not verified. |
| Output formats | Markdown, HTML, JSON, chunks, LaTeX, images. | HTML, Markdown, JSON, bounding boxes. |
| Input formats | PDF to Markdown converter. Other input types Not published. | PDF, PPT, Word and images. Excel is not supported in the open-source version. |
| Benchmarks | 83.7 on olmOCR-bench and 85.9 on OmniParseBench in vllm mode, both reported by Datalab. | Not published |
| Known weakness | Weights block free commercial use for companies above $5M. CPU fast mode drops to 66.6 on olmOCR-bench. | The open-source version uses community models, which are less accurate than the cloud's in-house models. |
Sources
What each one is
Marker
Hosted version of the open-source Marker and Surya models. Strong on books and papers.
Marker detailsChunkr
Document parsing API that segments and chunks pages for RAG. Self-hostable version available.
Chunkr detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "datalab/marker","fallbacks": ["chunkr/chunkr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Marker alternativesChunkr alternativesBest OCR models in 2026