Cyberax AI Playbook
cyberax.com
← All models
Model · IBM · Vision

Docling

An open-source toolkit from IBM Research that wraps layout, table and OCR models to turn PDFs, Word, PowerPoint, Excel, HTML, images and more into Markdown or lossless JSON for search and RAG. Now hosted by the LF AI & Data Foundation.

Modality
Vision
License
MIT (Open)
Released
July 15, 2024
Last verified
October 28, 2026
Runs locally
Yes

What you need to run it

Mac
Yes. Any Apple Silicon Mac (M1 or newer) with at least 8 GB (16 GB recommended) of memory. Newer chips are faster, but memory is what decides whether it runs.
Windows / Linux PC
NVIDIA works best. A graphics card with at least 4 GB (8 GB recommended) of video memory.
No graphics card
Works. Runs on a normal computer's processor, just slower.

Runs on a plain Mac or PC, and has an Apple-Silicon (MLX) mode for the Granite-Docling model.

Numbers are for: standard pipeline (CPU) or the 258M Granite-Docling model. Our estimate from the model’s size — the maker publishes no figure.

Strengths

  • MIT licence — free for any commercial use
  • Reads a very wide range of formats: PDF, DOCX, PPTX, XLSX, HTML, EPUB, email, images, LaTeX, even audio
  • Advanced PDF understanding: page layout, reading order, table structure, code and formulas
  • Optional vision-language pipeline using IBM's GraniteDocling-258M model
  • Plugs into AI tooling directly, including an MCP server for agents

Weaknesses

  • A toolkit, not a single model — quality depends on which pipeline and OCR engine you configure
  • Very frequent releases (v2.143 by October 2026), so pin a version in production

Try it

WhereTypeNotes
GitHub local pip install docling
Hugging Face (GraniteDocling) weights Optional VLM, Apache 2.0

Official sources

Change log

  • — Added hardware requirements (Mac support, memory, NVIDIA). Source: estimate from size and community builds (github docling-project/docling README (MPS/MLX section)).
  • — Initial entry. First public release v0.1.1 on 2024-07-15 (GitHub releases); latest is v2.143.1 (2026-10-27). MIT licence. Added 2026-10-28 from the popularity sweep: 69.7k GitHub stars.

Related models

MinerU

OpenDataLab · Vision

An open-source tool that wraps document-parsing models to turn PDFs, scans, Word and PowerPoint files into clean Markdown or JSON, keeping tables, formulas and reading order. Version 4.0 adds a local document library and agent-friendly reading with stable page and block citations.

MinerU Open Source License Open weights Runs locally Mac · 8 GB+
Try on: GitHub · Hugging Face
Released Sep 2026 · Verified 2026-10-28

OCR 4.1

Mistral AI · Vision

Mistral's current OCR service, powering its Document AI stack. Adds native paragraph-level bounding boxes, structural block labels, and block-level confidence scores over OCR 3.

Mistral Premier Proprietary
Try on: Mistral La Plateforme
Released Jul 2026 · Verified 2026-10-28

PaddleOCR 3.7

Baidu · Vision

Baidu's open-source OCR toolkit. Version 3.7 ships PP-OCRv6, a classic text-detection and recognition pipeline in tiny, small and medium sizes, alongside PaddleOCR-VL-1.6, a 0.9B vision-language model for full document parsing into Markdown or JSON.

Apache 2.0 Open Runs locally Mac · 8 GB+
Try on: GitHub · Hugging Face (PP-OCRv6) · Hugging Face (PaddleOCR-VL-1.6)
Released Jun 2026 · Verified 2026-10-28

GLM-OCR

Z.ai · Vision

Z.ai's small open OCR model for complex documents — under 1B parameters, built on the GLM-V architecture, and strong on tables, formulas, seals and code-heavy pages. Its SDK adds layout analysis for full document parsing.

MIT Open Runs locally Mac · 8 GB+
Try on: Hugging Face · Ollama · Z.ai API
Released Jan 2026 · Verified 2026-10-28

Llama 3.2 Vision

Meta · Vision

Meta's open-weight vision-language family. 11B and 90B variants — the practical open-weights vision model for self-hosted multimodal applications.

Llama 3.2 Community License Open weights Runs locally Mac · 16 GB+ 128k ctx
Try on: Hugging Face · OpenRouter · Ollama
Released Sep 2024 · Verified 2026-06-08