Marker vs Hunyuan OCR
Pick Marker to turn PDFs such as books and papers with equations into Markdown, JSON or RAG chunks, with a hosted API as an option. Pick Hunyuan OCR for photos, scene text, image translation and low-resource or historical scripts on a ~1B model. The main split is the task: Marker is a PDF converter, Hunyuan OCR is a general OCR VLM, and their licences limit different users.
Markerdatalab/marker | Hunyuan OCRtencent/hunyuan-ocr | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 14 credits / page | 2 credits / page |
| Per 1,000 pages | $14.00 | $2.00 |
| Typical speed | 2.5 s / page | 1.5 s / page |
| Best for | Books and research papers | Chinese documents and photos |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | 90+ languages | 100+ languages |
| Max pages per request | 1,000 | 500 |
| 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 self-hosted PDF to Markdown for books and papers with equations, with LaTeX output.
- You need RAG chunks or JSON block trees, not just text.
- You want a hosted option: Datalab's API includes about 5,000 free pages per month, and a CPU fast mode runs at 23.6 pages per second.
Choose Hunyuan OCR if
- You work with photos and scene text, text spotting, or text-image translation, and want bounding boxes in the output.
- You need low-resource or ancient-script OCR, which the vLLM client exposes among its 12 task types.
- You want a ~1B model that handles parsing, spotting, extraction and translation in one place, and that can run on a laptop through llama.cpp GGUF.
Key differences
| Marker | Hunyuan OCR | |
|---|---|---|
| Main job | PDF to Markdown converter built on the Surya OCR models | End-to-end OCR VLM for parsing, text spotting, extraction and image translation |
| Licence | Code Apache-2.0; weights under a modified AI Pubs Open Rail-M licence, free for research, personal use and startups under $5M | Tencent Hunyuan Community License; excludes EU, UK and South Korea, and needs a separate licence above 100M monthly active users |
| How you run it | Self-host on GPU, CPU or Apple Silicon, or use Datalab's hosted API | Self-host only, via vLLM, Transformers or llama.cpp |
| Hardware | vllm mode needs an NVIDIA GPU; fast mode runs on CPU with about 18 MB of ONNX weights | NVIDIA GPU with CUDA 13 recommended for the unified vLLM environment; llama.cpp GGUF for CPU, consumer GPU or laptop |
| Output formats | Markdown, HTML, JSON, chunks, LaTeX, images | Markdown, HTML, LaTeX, JSON, plain text, bounding boxes |
| Benchmark | olmOCR-bench 83.7 in vllm mode and 66.6 in CPU fast mode, reported by Datalab | ParseBench-2 overall 67.11 (v1.5), reported independently; a different benchmark, so the scores are not comparable |
| Known limits | Weights block free commercial use for companies above $5M; all README benchmarks are Datalab's own | Needs CUDA 13 and vLLM 0.25.1+ for the unified environment; an archived 1.0 note said Transformers and vLLM outputs differed in accuracy |
Sources
What each one is
Marker
Hosted version of the open-source Marker and Surya models. Strong on books and papers.
Marker detailsHunyuan OCR
Small end-to-end OCR model from Tencent covering parsing, extraction and photo translation.
Hunyuan OCR 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": ["tencent/hunyuan-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Marker alternativesHunyuan OCR alternativesBest OCR models in 2026