---
title: "Granite 4.0 3B Vision: Compact Multimodal Intelligence for Enterprise Documents"
description: "Granite 4.0 3B Vision is a compact vision-language model designed to extract information from enterprise documents, including tables, charts, forms, and semantically meaningful key-value pairs. It combines a million-scale ChartNet dataset with a code-guided synthesis pipeline, DeepStack Injection for routing semantic and high-resolution visual features to different layers, and a LoRA adapter architecture built on Granite 4.0 Micro. The modular design supports standalone image processing, text-only fallback, and integration with Docling for multi-page document pipelines. Reported evaluations show 86.4% on Chart2Summary, leading the evaluated models, second place on Chart2CSV at 62.1%, leading scores on the listed table benchmarks, and 85.5% zero-shot exact-match accuracy on VAREX."
---

# Granite 4.0 3B Vision: Compact Multimodal Intelligence for Enterprise Documents

[Hugging Face](https://yomu.fyi/company/hugging-face) · Madison Lee, Rogerio Feris, Eli Schwartz, Dhiraj Joshi, Pengyuan Li, Isaac Sanchez · Mar 31, 2026

**Type:** Announcement

## Summary

Granite 4.0 3B Vision is a compact vision-language model designed to extract information from enterprise documents, including tables, charts, forms, and semantically meaningful key-value pairs. It combines a million-scale ChartNet dataset with a code-guided synthesis pipeline, DeepStack Injection for routing semantic and high-resolution visual features to different layers, and a LoRA adapter architecture built on Granite 4.0 Micro. The modular design supports standalone image processing, text-only fallback, and integration with Docling for multi-page document pipelines. Reported evaluations show 86.4% on Chart2Summary, leading the evaluated models, second place on Chart2CSV at 62.1%, leading scores on the listed table benchmarks, and 85.5% zero-shot exact-match accuracy on VAREX.

## Context

Vision-language models can struggle with charts because interpretation requires reasoning over visual patterns, numerical data, natural language, and spatially precise details. Enterprise document extraction also depends on preserving layout relationships in tables, charts, forms, and other structured visuals.

## Approach / What changed

The model uses ChartNet, a 1.7 million-sample multimodal chart dataset with aligned plotting code, rendered images, data tables, summaries, and question-answer pairs. Its DeepStack Injection architecture routes abstract visual features to earlier layers and high-resolution spatial features to later layers. Granite 4.0 3B Vision is distributed as a LoRA adapter on Granite 4.0 Micro and can operate standalone or with Docling.

## Takeaways

- ChartNet contains 1.7 million chart samples spanning 24 chart types and six plotting libraries, with synthetic, human-annotated, and real-world subsets.
- DeepStack Injection separates semantic and spatial visual features across model layers, supporting layout-sensitive tasks such as table extraction, chart interpretation, and key-value parsing.
- The model achieved 92.1 on cropped PubTables-v2, 79.3 on full-page PubTables-v2, 64.0 on OmniDocBench, 88.1 on TableVQA, and 85.5% zero-shot exact match on VAREX.

**Tags:** [Architecture](https://yomu.fyi/topic/architecture), [Machine Learning](https://yomu.fyi/topic/machine-learning), [Open Source](https://yomu.fyi/topic/open-source), [Performance](https://yomu.fyi/topic/performance)

- Source: [Hugging Face](https://huggingface.co/blog/ibm-granite/granite-4-vision)
- Source URL: https://huggingface.co/blog/ibm-granite/granite-4-vision
- Ingested by Yomu: 2026-08-29T01:47:07.446Z

[Read original post](https://huggingface.co/blog/ibm-granite/granite-4-vision)