Surya
Datalab's open-source OCR toolkit for multilingual text recognition, layout analysis, reading order and table recognition.
About Surya
Scanned and complex documents need more than plain text extraction: line detection, layout blocks, reading order and tables across many languages.
Who it’s for
- Developers building document parsing or RAG pipelines they host themselves
- Researchers digitising multilingual scans and archives
When to consider it
Consider Surya when you want a self-hosted OCR and layout model that outputs tables as HTML and runs on an NVIDIA GPU (vLLM) or on CPU and Apple Silicon (llama.cpp); it is the engine behind Datalab's Marker converter.
Tradeoffs & limitations
- The code is Apache-2.0, but the model weights use a modified OpenRAIL-M licence: commercial use needs a paid licence if your company had more than $5M in gross revenue in the prior year or has raised more than $5M, unless use is limited to personal use or research.
- Layout, OCR and table recognition need a running inference backend (vLLM in Docker with the NVIDIA Container Toolkit, or llama.cpp's llama-server); Python 3.10 or newer is required.
SOTA overview · Documentation-based assessment · Sources & review method
Updates
No updates shared yet.
Discussion
Newest firstAsk a question or share how you use Surya.
Keep it helpful. Community rules
Loading discussion…
Sources & review method
Documentation-based assessment · Nov 1, 2026 · Prepared with AI assistance; not a hands-on benchmark.
Checked by SOTA · AI-assisted documentation review. Selection advice is our assessment; verify current requirements for your deployment.
Import history & original evidence
Surya on GitHub
Datalab's open-source OCR toolkit for multilingual text recognition, layout analysis, reading order and table recognition.
Based on official pages and announcements checked on 2026-11-01. No hands-on test, performance benchmark or popularity ranking is claimed.