Skip to content

InternVL

An open-source vision foundation model for vision-language tasks

Visit official site

What InternVL is

The model is described as an open-source vision foundation model built by scaling a vision transformer and aligning it with a language model. The model is listed with tasks ranging from visual perception and cross-modal retrieval to multimodal dialogue.

What you can do with it

  • Perform image classification
  • Run semantic segmentation
  • Classify video content
  • Retrieve images with text and build vision-language models

Example uses

  • Classify images with InternViT-6B
  • Retrieve image and text content with InternVL-C
  • Answer visual questions with InternVL-Chat

Model notes

  • The vision transformer is scaled to 6 billion parameters
  • The visual model is aligned with a language model
  • The result is described as a 14B open-source vision foundation model
  • The material reports 32 state-of-the-art results across wide-ranging tasks

Pros & cons

✓ What we like

  • A 6-billion-parameter vision transformer aligned with a language model
  • Named variants for classification, retrieval and visual question answering

! What to watch out for

  • The named use cases depend on specific model variants
  • Vision tasks are paired with a language model for alignment

FAQ

How large is the model?

The material describes a 6-billion-parameter vision transformer aligned into a 14B open-source vision foundation model.

Who is it for?

The listed audience is computer vision research and multimodal application development.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review