AI Tools.

Search
image text to text by ibm-granite

granite-docling-258M

granite-docling-258M is a 258M-parameter vision-language model fine-tuned specifically for document understanding tasks within the Docling pipeline. It handles OCR, layout parsing, table extraction, formula recognition, and chart reading in a single inference pass. The model is built on the Idefics3 architecture and integrates directly with the open-source Docling library.

Summary text generated by an automated pipeline from the model card · Not individually reviewed or run by us · How this page is made

Share: X Reddit HN LinkedIn

From the model card

Fields below are copied from the tags and counters on the HuggingFace repository ibm-granite/granite-docling-258M at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.

Publisher (HF namespace)
ibm-granite
Pipeline tag
image-text-to-text
Library
Transformers
Weight formats
safetensors
License tag
apache-2.0 — read the license file in the repo before relying on it
Language tags
English (en)
Papers cited
arXiv:2501.17887, arXiv:2503.11576, arXiv:2305.03393
Datasets declared
ds4sd/SynthCodeNet, ds4sd/SynthFormulaNet, ds4sd/SynthChartNet, HuggingFaceM4/DoclingMatix
Downloads (HF counter at last fetch)
407,501
Likes (HF counter at last fetch)
1,254
Model card
https://huggingface.co/ibm-granite/granite-docling-258M

Use cases

  • Extracting structured tables from scanned PDFs
  • Parsing multi-column academic paper layouts
  • Recognising LaTeX formulas in technical documents
  • Converting chart images to underlying data
  • Building document ingestion pipelines for RAG systems

Pros

  • Tightly integrated with the Docling open-source ecosystem
  • Handles tables, formulas, charts, and OCR in one model
  • Apache 2.0 license; Azure deployment supported
  • 258M parameters make it deployable on modest hardware

Cons

  • Accuracy drops significantly on handwritten or degraded scans
  • Formula recognition is weaker than dedicated TeX parsers
  • No support for right-to-left script layouts
  • Performance linked to Docling pipeline version; breaking changes possible

Tags

transformerssafetensorsidefics3image-text-to-texttext-generationdocumentscodeformulachartocrlayouttabledocument-parsedoclinggraniteextractionmathconversationalendataset:ds4sd/SynthCodeNet