Granite Vision 3.3
Compact open vision-language model for enterprise document understanding.
About
Granite Vision 3.3 is IBM's compact open-source vision-language model, released in June 2025 as the 2B-parameter flagship of the 3.3 generation. It was trained by fine-tuning a Granite language model with both image and text modalities on instruction-following data curated for document understanding.
The model is purpose-built for visual document tasks: extracting content from tables, charts, infographics, plots, and diagrams. Experimental capabilities added in 3.3 include image segmentation, doctag generation, and multi-page document support. IBM reports enhanced safety alignment compared with earlier Granite vision models.
Released under Apache 2.0, it is designed for enterprise deployments that need on-device or self-hosted multimodal understanding. The open weights and small footprint make it a practical choice for OCR-style pipelines, RAG over scanned documents, and structured extraction from PDFs and slide decks.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 99; Trending score 74/100 from researcher assessment.