Surya vs Chunkr
Pick Surya if you want a small local model (689M parameters) that returns layout labels, reading order, OCR and HTML tables with bounding boxes for 90+ languages, and your company is under $5M in funding or revenue. Pick Chunkr if you want a parsing service for PDF, PPT, Word and images with Markdown output for RAG chunking, and you want the choice of a hosted API or self-hosting. The main split is scope: Surya is a model you run against your own inference server, while Chunkr is a parsing pipeline that also needs an LLM endpoint when self-hosted.
Suryadatalab/surya | Chunkrchunkr/chunkr | |
|---|---|---|
| Type | Open source | OCR API |
| Price | 2 credits / page | 14 credits / page |
| Per 1,000 pages | $2.00 | $14.00 |
| Typical speed | 1.5 s / page | 3.0 s / page |
| Best for | Layout and line-level OCR | Layout-aware chunks for RAG |
| Inputs | PDF, images | PDF, images, DOCX |
| Languages | 90+ languages | Multilingual |
| Max pages per request | 500 | 1,000 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | No | ✓ 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 Surya if
- You need to run OCR on a laptop: it runs on CPU or Apple Silicon via llama.cpp, with peak memory of about 4.3-5.9 GB on an M4 Max.
- You want line and block level bounding boxes, layout labels and reading order from one model, with HTML tables and KaTeX math.
- You want a published accuracy number: the authors report 83.7% on olmOCR-bench for Surya 2.5, and 4.52 pages/s at 192 DPI on one H100 with vLLM.
Choose Chunkr if
- You want RAG chunking of PDFs and slides, and input that includes PDF, PPT, Word and images.
- You want Markdown, HTML and JSON output with bounding boxes, from a service you can run in the cloud or self-host with Docker (GPU, CPU-only or Mac ARM).
- You already have an OpenAI-compatible LLM endpoint to plug in, and you can accept AGPL-3.0 or buy a commercial license.
Key differences
| Surya | Chunkr | |
|---|---|---|
| What it is | OCR, layout analysis, reading order and table recognition toolkit from Datalab | Document parsing service from Lumina AI with layout analysis, OCR and semantic chunking |
| Deployment | Self-host only: a lightweight client plus a vLLM or llama.cpp inference server | Hosted cloud API or open-source self-host; self-host needs an external LLM endpoint |
| Licence | Code Apache-2.0; weights under modified AI Pubs Open Rail-M, free for research, personal use and companies under $5M funding or revenue | AGPL-3.0 or commercial license; closed-source use needs the commercial license |
| Price | Free self-hosted (you pay for compute); commercial license needed above $5M revenue or funding | Hosted credit-based plans, prices not published; open-source version is free (you pay for compute and an LLM endpoint) |
| Output formats | JSON, HTML, LaTeX, bounding boxes, plain text | HTML, Markdown, JSON, bounding boxes |
| Benchmarks | olmOCR-bench 83.7% (Surya 2.5, author-reported); 74.9% pass rate across 91 languages | Not published |
| Known limits | Surya 2.5 tables lack row/column geometry; multilingual tables score 58.1; not for handwriting or photos | Open-source version uses community models, less accurate than the cloud's; no Excel in the open-source version |
Sources
What each one is
Surya
Open-source OCR toolkit with line detection, layout analysis, reading order and tables.
Surya 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/surya","fallbacks": ["chunkr/chunkr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Surya alternativesChunkr alternativesBest OCR models in 2026