> ## Documentation Index
> Fetch the complete documentation index at: https://docs.somark.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# SoMark vs. OXR

> Compare SoMark and OXR by parsing speed, accuracy, document recognition, and integration options.

**SoMark** is a production document parsing service powered by a separate model. It offers higher accuracy and more detailed recognition of document elements and structures.

**OXR** is an open-source model from the same team that parses PDFs and common images. You can try it through the free API or run it on your own infrastructure. Both produce Markdown and JSON.

## Parsing speed

| Deployment option | Parsing speed |
| - | - |
| Self-hosted | Under equivalent parsing conditions, a private SoMark deployment parses documents faster even when self-hosted OXR uses multiple GPUs. |
| SoMark API | Parses a 100-page document in as little as **5 seconds**. |

## Parsing capabilities

| Capability | SoMark | OXR |
| - | - | - |
| Document elements | **21 types** Title, text block, header, footer, footnote, figure, figure caption, table, table caption, equation, chemical structure, chemical equation, code block, sidebar, table of contents, table of contents entry, choice item, fill-in-the-blank, reference, QR code, stamp | **13 types** Title, body text, header, footer, footnote, figure, figure or table caption, table, formula, chemical structure, chemical equation, code block, stamp |
| Structure | Merges tables and text across pages, recognizes heading levels, and reduces interference from watermarks and stamps | Preserves layout coordinates and reading order |
| Image handling | Recovers images embedded in text and tables, recognizes text within images, and generates image descriptions | Detects image regions and returns the images |
| File formats | **26 formats** `pdf` `doc` `docx` `ppt` `pptx` `xls` `xlsx` `xlsm` `png` `jpg` `jpeg` `bmp` `tiff` `jp2` `dib` `ppm` `pgm` `pbm` `gif` `heic` `heif` `webp` `xpm` `tga` `dds` `xbm` | `pdf` `png` `jpg` `jpeg` `bmp` `tif` `tiff` `webp` |

## Ways to use each product

| Option | SoMark | OXR |
| - | - | - |
| Online demo | Upload a PDF, image, or Office file in the [SoMark demo](https://somark.ai/playground) | Upload a PDF or common image file in the [OXR demo](https://free-oxr.somark.ai) |
| Hosted API | The [SoMark API](https://docs.somark.ai/en/api-reference/index.md) requires an API key | The [Free OXR API](https://docs.somark.ai/en/open-source-tools/oxr.md#oxr-free-api) is available without registering or requesting an API key. Requests are queued, making it suitable for testing |
| Hosted API limits | **200 MB and 300 pages** per request | **10 MB and 20 pages** per request |
| Free usage | Up to **500 pages per day** and **2,000 pages per month**. Usage beyond either cap draws from paid quota. [SoMark billing rules](https://docs.somark.ai/en/documentation/free-tier-rules.md) | The Free OXR API is free to use |
| Deployment | Commercial private deployment with customization for your workflow | Get the code and model from [GitHub](https://github.com/SoMarkAI/OXR) to run OXR offline, with no fixed file size or page limit |

Both provide an HTTP API. You can reuse the [SoMark CLI and SDK](https://docs.somark.ai/en/cli-sdk/index.md) and configure the URL and authentication for the service you use. See [SoMark and OXR integration differences](https://docs.somark.ai/en/open-source-tools/oxr/integration) for request parameter and response differences.

## Which should you choose?

* **Choose SoMark** for production workflows that need fast, accurate parsing and detailed recognition of complex document structures. Use your own documents to evaluate the results. [Try SoMark online](https://somark.ai/playground)
* **Choose OXR** to test PDF or common image parsing without signing up, or to self-host and extend the open-source model.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.