selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

Alternatives to surya-ocr-2.5

120 open-weight models that do the same job as surya-ocr-2.5 (689M OCR), ranked by how close they are in task and size, then by popularity and how recent they are. Licence and VRAM are shown for each.

the original
datalab-to logo

surya-ocr-2.5

datalab-to·OCR

≈2.0 GB

We also ship smaller models for line-level text detection and ocr error detection. It works on a range of documents (see usage and benchmarks).

689M256K ctx OpenRAIL
144.4KOct 2026gguf
--vram
any
--gpus
--quant
  1. openbmb logo

    MiniCPM-OCR-Jev

    openbmb·OCR

    ≈5.5 GB

    MiniCPM-OCR Jev (2B) is an OCR quality-assessment model for text, formula, and table regions. Given a region image and a candidate OCR result, it returns a score from 0 to 1 — higher means the candidate is more likely correct.

    2.2B256K ctx Apache 2.0
    79.5KOct 2026gguf
    • Same category
    • Larger (2.2B vs 689M)
    • Commercial use OK
    vs
  2. lightonai logo≈2.8 GB

    Merged variant for extra robustness (advanced). This model combines the base checkpoint with RLVR-trained weights using task arithmetic merging, providing improved consistency across challenging document categories.

    1B16K ctx Apache 2.0
    379Jan 2026gguf
    • Same category
    • Similar size (1B vs 689M)
    • Commercial use OK
    vs
  3. microsoft logo

    Florence-2-base

    microsoft·OCR

    ≈1.0 GB

    This Hub repository contains a HuggingFace's transformers implementation of Florence-2 model from Microsoft.

    232M1K ctx MIT
    21.3MJun 2024
    • Same category
    • Smaller (232M vs 689M)
    • Commercial use OK
    • Widely used
    vs
  4. baidu logo

    Unlimited-OCR

    baidu·OCR

    ≈4.5 GB

    Please refer to the official vLLM recipe for deployment details:

    3.3BA1.1B32K ctx MIT
    5.6MJun 2026gguf
    • Same category
    • Larger (3.3B vs 689M)
    • Commercial use OK
    • Widely used
    vs
  5. HuggingFaceM4 logo

    Florence-2-DocVQA

    HuggingFaceM4·OCR

    ≈2.3 GB

    This is the model card of a 🤗 transformers model that has been pushed on the Hub. It has been automatically generated.

    823M1K ctx Unknown licence
    3.9KJun 2024
    • Same category
    • Similar size (823M vs 689M)
    vs
  6. ibm-granite logo

    granite-docling-258M

    ibm-granite·OCR

    ≈1.1 GB

    Granite Docling is a multimodal Image-Text-to-Text model engineered for efficient document conversion. It preserves the core features of Docling while maintaining seamless integration with DoclingDocuments to ensure full compatibility.

    258M8K ctx Apache 2.0
    423.9KMay 2025gguf
    • Same category
    • Smaller (258M vs 689M)
    • Commercial use OK
    vs
  7. baidu logo

    Unlimited-OCR-2

    baidu·OCR

    ≈4.5 GB

    Sparse Reference Sliding Window Attention (SR-SWA): Compresses text-visual attention interactions from ultra-long sequences down to a fixed-size set of working references (e.g., 512 tokens).

    3.3BA1.1B128K ctx MIT
    621.8KOct 2026gguf
    • Same category
    • Larger (3.3B vs 689M)
    • Commercial use OK
    • Newer release
    vs
  8. ByteDance logo

    Dolphin-1.5

    ByteDance·OCR

    ≈1.4 GB

    Dolphin (Document Image Parsing via Heterogeneous Anchor Prompting) is a novel multimodal document image parsing model that follows an analyze-then-parse paradigm.

    398M4K ctx MIT
    529Oct 2025
    • Same category
    • Similar size (398M vs 689M)
    • Commercial use OK
    vs
  9. tiiuae logo

    Falcon-OCR

    tiiuae·OCR

    ≈1.1 GB

    Falcon OCR is a 270M parameter early-fusion vision-language model for document OCR. Given an image, it can produce plain text, LaTeX for formulas, or HTML for tables.

    270M16K ctx Apache 2.0
    5.5KFeb 2026
    • Same category
    • Smaller (270M vs 689M)
    • Commercial use OK
    vs
  10. ByteDance logo

    Dolphin

    ByteDance·OCR

    ≈1.4 GB

    Dolphin (Document Image Parsing via Heterogeneous Anchor Prompting) is a novel multimodal document image parsing model that follows an analyze-then-parse paradigm.

    398M4K ctx MIT
    670May 2025
    • Same category
    • Similar size (398M vs 689M)
    • Commercial use OK
    vs
  11. deepseek-ai logo

    DeepSeek-OCR

    deepseek-ai·OCR

    ≈4.5 GB

    Refer to 🌟GitHub for guidance on model inference acceleration and PDF processing, etc.

    3.3BA1.1B8K ctx MIT
    1.8MOct 2025ollama
    • Same category
    • Larger (3.3B vs 689M)
    • Commercial use OK
    • Widely used
    vs
  12. echo840 logo

    MonkeyOCR-pro-1.2B

    echo840·OCR

    ≈3.2 GB

    MonkeyOCR: Document Parsing with a Structure-Recognition-Relation Triplet Paradigm Zhang Li, Yuliang Liu, Qiang Liu, Zhiyin Ma, Ziyang Zhang, Shuo Zhang, Zidun Guo, Jiarui Zhang, Xinyu Wang, Xiang Bai

    1.2BUnknown licence
    125Jul 2025
    • Same category
    • Similar size (1.2B vs 689M)
    vs
  13. dots-studio logo

    dots.mocr

    dots-studio·OCR

    ≈4.1 GB

    We present dots.mocr. Beyond achieving state-of-the-art (SOTA) performance in standard multilingual document parsing among models of comparable size, dots.mocr excels at converting structured graphics (e.g., charts, UI layouts, scientific figures and etc.) directly into SVG code.

    3B128K ctx MIT
    57.4KMar 2026gguf
    • Same category
    • Larger (3B vs 689M)
    • Commercial use OK
    vs
  14. dots-studio logo

    dots.ocr

    dots-studio·OCR

    ≈4.1 GB

    dots.ocr is a powerful, multilingual document parser that unifies layout detection and content recognition within a single vision-language model while maintaining good reading order. Despite its compact 1.7B-parameter LLM foundation, it achieves state-of-the-art(SOTA) performance.

    3B128K ctx MIT
    337.2KJul 2025gguf
    • Same category
    • Larger (3B vs 689M)
    • Commercial use OK
    vs
  15. vikp logo

    texify

    vikp·OCR

    ≈1.2 GB

    texify is an open-weight ocr model with 312.5M parameters from vikp, published on Hugging Face.

    312M2K ctx CC BY-SA 4.0
    48.2KDec 2023
    • Same category
    • Smaller (312M vs 689M)
    • Commercial use OK
    vs
  16. deepseek-ai logo

    DeepSeek-OCR-2

    deepseek-ai·OCR

    ≈4.5 GB

    Refer to 🌟GitHub for guidance on model inference acceleration and PDF processing, etc.

    3.4BA1.2B8K ctx Apache 2.0
    183.6KJan 2026gguf
    • Same category
    • Larger (3.4B vs 689M)
    • Commercial use OK
    vs
  17. microsoft logo

    Florence-2-base-ft

    microsoft·OCR

    ≈1.0 GB

    This Hub repository contains a HuggingFace's transformers implementation of Florence-2 model from Microsoft.

    232M1K ctx MIT
    156.3KJun 2024
    • Same category
    • Smaller (232M vs 689M)
    • Commercial use OK
    vs
  18. nanonets logo

    Nanonets-OCR2-3B

    nanonets·OCR

    ≈5.0 GB

    Nanonets-OCR2 by Nanonets is a family of powerful, state-of-the-art image-to-markdown OCR models that go far beyond traditional text extraction.

    3.8B125K ctx Unknown licence
    892.7KOct 2025gguf
    • Same category
    • Larger (3.8B vs 689M)
    vs
  19. PaddlePaddle logo

    PP-OCRv6_medium_det

    PaddlePaddle·OCR

    ? GB

    PP-OCRv6 is a lightweight OCR system that combines architectural innovation with data-centric optimization. It redesigns the backbone, detection neck, and recognition neck around a unified MetaFormer-style building block with structural reparameterization.

    ?BApache 2.0
    47.4KJun 2026
    • Same category
    • Commercial use OK
    vs
  20. PaddlePaddle logo

    PP-OCRv5_server_det

    PaddlePaddle·OCR

    ? GB

    PP-OCRv5_server_det is one of the PP-OCRv5_det series, the latest generation of text detection models developed by the PaddleOCR team.

    ?BApache 2.0
    469.8KJun 2025
    • Same category
    • Commercial use OK
    vs
  21. nanonets logo≈4.2 GB

    Nanonets-OCR2 by Nanonets is a family of powerful, state-of-the-art image-to-markdown OCR models that go far beyond traditional text extraction.

    1.6B32K ctx Apache 2.0
    176Oct 2025gguf
    • Same category
    • Larger (1.6B vs 689M)
    • Commercial use OK
    vs
  22. PaddlePaddle logo

    en_PP-OCRv5_mobile_rec

    PaddlePaddle·OCR

    ? GB

    en_PP-OCRv5_mobile_rec is one of the PP-OCRv5_rec that are the latest generation text line recognition models developed by PaddleOCR team. It aims to efficiently and accurately support the recognition of English. The key accuracy metrics are as follow:

    ?BApache 2.0
    202.6KAug 2025
    • Same category
    • Commercial use OK
    vs
  23. datalab-to logo

    chandra-ocr-2

    datalab-to·OCR

    ≈6.8 GB

    Chandra 2 is a state of the art OCR model from Datalab that outputs markdown, HTML, and JSON. It is highly accurate at extracting text from images and PDFs, while preserving layout information.

    5.3B256K ctx OpenRAIL
    1.3MMar 2026gguf
    • Same category
    • Larger (5.3B vs 689M)
    vs
  24. PaddlePaddle logo

    PP-OCRv6_small_det

    PaddlePaddle·OCR

    ? GB

    PP-OCRv6 is a lightweight OCR system that combines architectural innovation with data-centric optimization. It redesigns the backbone, detection neck, and recognition neck around a unified MetaFormer-style building block with structural reparameterization.

    ?BApache 2.0
    12.8KJun 2026
    • Same category
    • Commercial use OK
    vs