Surya is an open-source document OCR toolkit for extracting text and document structure across many languages. It can recognize page text, detect text lines, identify layout regions such as tables and images, and— in Surya 2—detect reading order and recognize tables. You can run it yourself or use Datalab’s hosted platform. The right choice depends on your language and document mix, deployment needs, and licensing requirements.
What Surya OCR does
Surya is designed for document pages rather than only clean, isolated text. The project describes OCR support for more than 90 languages, line-level text detection in any language, and layout analysis for regions including tables, images, and headers. Example document types include scanned pages, forms, textbooks, newspapers, scientific papers, and presentations.
Datalab describes Surya 2 as a 650-million-parameter model combining four tasks: full-page OCR, layout analysis, reading-order detection, and table recognition. Line detection is described separately in the project’s training notes, so it should not be confused with those four Surya 2 tasks.
How accurate and fast is it?
Datalab reports an 83.3% result on olmOCR-bench. Its internal evaluation across 91 languages reports an 87.2% pass rate overall, with 38 languages scoring at least 90%. These are publisher-reported figures, not independent tests, and an aggregate language score does not establish performance for every language, script, scan quality, handwriting style, or document layout.
#1 Best Overall
- PORTABLE SCANNER FOR USE ON-THE-GO — The fastest and lightest mobile single-sheet-fed compact document scanner in its class¹
- QUICK DOCUMENT SCANNING ― This Epson ultra-fast scanner scans a single page as quickly as 5.5 seconds²; Windows and Mac compatible
- VERSATILE PAPER HANDLING ― Portable scanner scans documents up to 8.5 x 72 in; Also easily digitizes receipts and ID cards to make accounting, bookkeeping, and organizing simpler
- INTUITIVE, HIGH-SPEED SOFTWARE — Epson ScanSmart Software³ is a smart tool allowing you to easily scan, review, and save; Stay organized easily with the help of this Epson scanner
- EASY SETUP — USB-powered connect to your computer for quick and simple scanning; No batteries or external power supply required to operate portable document scanner; Standard Connectivity: USB 2.0
Datalab also reports throughput of 5.35 pages per second on one RTX 5090 at 128 concurrent requests. Treat that as a result under the stated GPU and concurrency conditions, not a general speed guarantee for a different machine or workload.
The published figures do not provide an apples-to-apples comparison with Tesseract or cloud OCR services. To compare systems for your use case, run the same representative documents through each option and assess text accuracy, structure, latency, and operating cost.
Rank #2
- FAST SPEEDS - Scans color and black and white documents a blazing speed up to 16ppm (1). Color scanning won’t slow you down as the color scan speed is the same as the black and white scan speed.
- ULTRA COMPACT – At less than 1 foot in length and only about 1. 5lbs in weight you can fit this device virtually anywhere (a bag, a purse, even a pocket).
- READY WHENEVER YOU ARE – The DS-640 mobile scanner is powered via an included micro USB 3. 0 cable allowing you to use it even where there is no outlet available. Plug it into you PC or laptop and you are ready to scan.
- WORKS YOUR WAY – Use the Brother free iPrint&Scan desktop app for scanning to multiple “Scan-to” destinations like PC, Network, cloud services, Email and OCR. (2) Supports Windows, Mac and Linux and TWAIN/WIA for PC/ICA for Mac/SANE drivers. (3)
- OPTIMIZE IMAGES AND TEXT – Automatic color detection/adjustment, image rotation (PC only), bleed through prevention/background removal, text enhancement, color drop to enhance scans. Software suite includes document management and OCR software. (4)
How to install and run Surya
Repository setup
- Check the environment: the repository documents Python 3.9 or later and PyTorch.
- Install the package: run
pip install surya-ocr. - Run an initial job: model weights download automatically on first use. The repository documents device auto-detection and optional use of a Streamlit GUI.
The documented repository setup pins Transformers 4.36.2; that repository snapshot says Transformers 4.37 and newer were not supported. This is a compatibility note for that documented setup, not a guarantee about later package releases. Check the current repository instructions before changing dependencies.
Surya 2 serving backends
Datalab describes vLLM for NVIDIA GPU deployment and llama.cpp for CPU or Apple Silicon. NVIDIA deployments may require Docker and the NVIDIA Container Toolkit; CPU and Apple Silicon users need a llama.cpp llama-server binary. These Surya 2 serving requirements are distinct from the repository’s general package-install instructions, so confirm the requirements for the particular release and backend you plan to use.
Recommended Free Tools
Rank #3
- FAST DOCUMENT SCANNING — Document scanner with feeder allows you to speed through stacks with a 50-sheet Auto Document Feeder (ADF); Efficient office scanner to help you scan more productively
- INTUITIVE, HIGH-SPEED SOFTWARE — Quickly scan with this desktop document scanner; Epson ScanSmart Software lets you easily preview scans, email files, upload to the cloud, and more; Plus, automatic file naming saves even more time
- SEAMLESS INTEGRATION — Easily incorporate your data into most document management software with the included TWAIN driver; Office document scanner integrates seamlessly with business workflows
- EASY SHARING — Duplex scanner allows you to scan straight to email or popular cloud storage2 services like Dropbox, Evernote, Google Drive, and OneDrive for simple storage and sharing
- SIMPLE FILE MANAGEMENT — Scanner allows the creation of searchable PDFs with Optical Character Recognition (OCR) and convert scans to editable Word or Excel files effortlessly; Designed for home and office document scanning
Can Surya run without a GPU?
Local CPU use is an option: Datalab describes llama.cpp for CPU deployments, as well as for Apple Silicon. The repository also documents automatic device detection. A GPU is therefore not an absolute requirement for every local setup, but no CPU speed figure is provided. Test your expected page volume and latency on the hardware you intend to operate.
Can Surya extract tables from PDFs?
Surya 2 includes table recognition, and Surya’s layout analysis identifies table regions. That makes it a candidate for documents where table structure matters, but it does not guarantee that every PDF table will be reconstructed correctly. PDF source quality, complex or merged cells, and surrounding layout can affect the result. Verify the extracted rows, columns, and values against the original pages before relying on them downstream.
Rank #4
- Scanner type: Document
- Connectivity technology: USB
- With Auto Scan Mode, the scanner automatically detects what you're scanning
- Digitize documents and images
Choose self-hosted Surya or Datalab’s platform
| Option | What it offers | What you take on |
|---|---|---|
| Self-hosted | Local processing on CPU, Apple Silicon, or NVIDIA GPU; can suit privacy-sensitive or offline workflows. | You manage dependencies, suitable hardware, model serving, scaling, and maintenance. |
| Datalab hosted platform | Hosted API and SDK, versioned pipelines, managed cloud, EU data residency, VPC deployment, and on-premises or air-gapped options. | Evaluate the service’s fit for your data, workflow, operating requirements, and budget. |
Compare both approaches using the same four criteria: accuracy on your language and layout mix, latency and throughput, data residency and compliance, and total operating cost. For self-hosting, include hardware availability and the work of running and maintaining the model.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What Datalab’s published plans cost
| Plan | Published price or allowance |
|---|---|
| Free | Allowance described as $20 per month with a work email or $10 per month with a personal email. |
| Team | $400 per month. |
| Enterprise | Custom pricing. |
These are the plans and allowances stated on Datalab’s pricing page in the available information; confirm current terms directly with Datalab before budgeting.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsBest Value
- OUR MOST ADVANCED SCANSNAP. Large touchscreen, fast 45ppm double-sided scanning, 100-sheet document feeder, Wi-Fi and USB connectivity, automatic optimizations, and support for cloud services. Upgraded replacement for the discontinued iX1600
- CUSTOMIZABLE. SHARABLE. Select personalized profiles from the touchscreen. Send to PC, Mac, mobile devices, and clouds. QUICK MENU lets you quickly scan-drag-drop to your favorite computer apps
- STABLE WIRELESS OR USB CONNECTION. Built-in Wi-Fi 6 for the fastest and most secure scanning. Connect to smart devices or cloud services without a computer. USB-C connection also available
- PHOTO AND DOCUMENT ORGANIZATION MADE EFFORTLESS. Easily manage, edit, and use scanned data from documents, receipts, photos, and business cards. Automatically optimize, name, and sort files
- AVOIDS PAPER JAMS AND DAMAGE. Features a brake roller system to feed paper smoothly, a multi-feed sensor that detects pages stuck together, and skew detection to prevent paper damage and data loss
Licensing: code and model weights differ
Surya’s code is licensed under Apache 2.0, while its model weights use a modified OpenRAIL-M license. Datalab’s announcement describes free use for research, personal projects, and startups under $5 million. Its pricing page separately says licensing depends on funding or revenue thresholds and directs larger commercial users toward licensing or hosted or on-premises arrangements. Because code and weights have different terms, do not assume that an Apache-licensed codebase makes every commercial use of the weights unrestricted. Check the current weight license and obtain written guidance from Datalab for a business outside the stated thresholds.
How to evaluate Surya for production
Build a test set that reflects the documents the system will actually process, rather than relying only on aggregate benchmark scores. Include the languages, scripts, scan conditions, handwriting (if relevant), table designs, and document domains in your workflow.
Quick Recap
- Measure character or word accuracy for extracted text.
- Check whether reading order matches the page, especially in multi-column layouts.
- Compare detected layout regions and table structures with the source documents.
- Evaluate downstream extraction quality: whether the recognized content is usable in the next step of your workflow.
- Measure latency and throughput on the intended hardware or hosted plan at realistic concurrency.
- Review dependency compatibility, backend requirements, data residency, and license terms for the release you intend to deploy.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
