SU

surya ocr 2

SuryaGGUF by datalab-toUpdated 2026-05-27
Download on Hugging Face
Parameters632M
Context256K
Layers24
Architectureqwen35
Smallest quant1.3 GB
Quick answer

surya ocr 2 needs about 2.2 GB of memory at its main file (a 1.3 GB file) with an 8K context window, so it fits on a 4 GB GPU. Longer context needs more: at 32K tokens the KV cache alone adds about 1.6 GB.

Surya OCR 2 – 650 M param OCR model for document layout, text, and table recognition

Surya OCR 2 is a 0.632 B param model built on qwen35 architecture. It detects layout, performs OCR, and recognizes tables in a single pass. The model supports 91 languages and achieves high accuracy on internal benchmarks.

Good at

  • Layout analysis with reading order
  • OCR of diverse documents
  • Table recognition (rows & columns)

Limits

  • Requires 1.27 GB GGUF file, ~2.2 GB VRAM at 8 k context
  • No vision support beyond OCR
  • OpenRail license limits commercial use without payment

Best for

  • Processing scanned PDFs for data extraction
  • Automated form reading
  • Table extraction from invoices

Who it’s for. Developers building document intelligence pipelines who need a single model for layout, OCR, and table detection. Suitable for research, startups, and small‑scale commercial use under the OpenRail license.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—1.3 GB 2.2 GB
Fits Get file →

The only available quant is the default 1.27 GB file. It needs approximately 2.2 GB VRAM at 8 k context. For lower VRAM, consider using llama.cpp with the same file, but performance will be slower.

What GPU do you need for surya ocr 2?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is the main file (1.3 GB file, about 2.2 GB in use).

Specifications

Architecture from the GGUF header

Architectureqwen35
Parameters632M
Layers24
Trained context262,144 tokens
KV cache (fp16)48 KB per token

Source

Quantized bydatalab-to
FamilySurya
Licenseopenrail
Downloads (HF)557,195
Last updated2026-05-27

Frequently asked questions about surya ocr 2

How much VRAM does surya ocr 2 need?

About 2.2 GB at its main file with an 8K context window. That is the 1.3 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run surya ocr 2 on an 8 GB GPU?

Yes, with the smaller files: the 1.3 GB file needs about 2.2 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of surya ocr 2 GGUF?

It depends on the quant: file is 1.3 GB.

What context length does surya ocr 2 support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.4 GB at 8K and 1.6 GB at 32K.

Where can I download surya ocr 2 GGUF?

The files are on Hugging Face in the datalab-to/surya-ocr-2-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does surya ocr 2 use?

Hugging Face lists the license as openrail. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Surya OCR 2 on 8 GB VRAM?

Yes, the model requires about 2.2 GB VRAM at 8 k context, so 8 GB is sufficient.

Does Surya OCR 2 support image input?

Surya OCR 2 processes images and PDFs; it performs layout, OCR, and table detection on visual documents.

What license applies to Surya OCR 2?

The model weights are under a modified AI Pubs Open Rail-M license, free for research, personal use, and startups under $5M.

Similar models

Browse all →

More Surya models

All Surya →

Other tiny models

All tiny models →