docling Releases

This release introduces an experimental TableCropsLayoutModel and a factory/plugin-capability system for Layout and Table models. It includes fixes for input formats, XLSX singleton cells, and DOCX list items.

Dec 2, 2025

The update adds save and load functionality for conversion results and fixes issues with threaded pipelines, document timeouts, and GPU availability for RapidOCR. It also updates documentation for various integrations.

Nov 20, 2025

This release introduces an Image backend and an experimental Layout + VLM model with layout prompt. It also fixes model repository names and adds DOCX header/footer parsing.

Nov 17, 2025

A minor patch release that defaults to EasyOCR in Python 3.14 to address compatibility issues.

Nov 10, 2025

This is a maintenance release addressing performance bottlenecks in table parsing for DOCX and HTML formats, alongside improvements to documentation navigation.

Nov 6, 2025

This release introduces tracking capabilities for VLM model generation details and applies stability fixes to NuExtract and OCR parameter handling.

Nov 6, 2025

This is a minor patch release focusing on correcting image request response extraction logic in picture descriptions.

Nov 4, 2025

Introduces threading support in the standard pipeline while maintaining legacy behavior, alongside PDF threading fixes and documentation updates.

Oct 31, 2025

Adds support for Python 3.14 and exposes token counts in VLM predictions, alongside improvements to Excel parsing and documentation.

Oct 30, 2025

Adds password protection support for PDFs, MLX Whisper for Apple Silicon, and performance improvements by defaulting to docling-parse-v4.

Oct 22, 2025

Focuses on DOCX processing (DrawingML) and fixes page concatenation issues in the VLM pipeline, with added documentation for PII handling.

Oct 15, 2025

This is a small patch release optimizing model downloads by skipping EasyOCR by default.

Oct 13, 2025

Introduces auto-OCR model selection and Tesseract PSM options, alongside numerous fixes for ASR, Excel chartsheets, and table handling.

Oct 13, 2025

A maintenance release addressing Markdown rendering issues, improving table handling, and updating documentation with new examples.

Oct 3, 2025

Introduces rich tables for the HTML backend and refines the GraniteDocling model with new stopping criteria and updated inference tools.

Sep 30, 2025

Enhances MSWord processing with rich tables and adds a new backend parser for WebVTT files, alongside various bug fixes.

Sep 22, 2025

Deploys the new granite-docling model and enhances RapidOCR capabilities with support for generic extra arguments.

Sep 17, 2025

Enables enrichment steps across all conversion pipelines and provides new documentation for RAG and MCP integration.

Sep 11, 2025

Optimizes performance by updating default parameters and the backend for the new docling-parse library.

Sep 5, 2025

Switches the default layout model to Heron, replacing the previous model.

Sep 3, 2025