Skip to content

InternVL2_5-4B-MPO

A multimodal model built with preference optimisation

Visit official site

What InternVL2_5-4B-MPO is

The model is described as a multimodal large language model series built from InternVL2.5 and mixed preference optimisation, joining a vision encoder with pretrained language models.

What you can do with it

  • Generate captions from images and video data
  • Produce automatic annotations and summaries for video

Model setup

  • Install Transformers and Torch before loading the model
  • Preprocess images by resizing and converting their format

Pros & cons

✓ What we like

  • Multi-image and video data are supported
  • Architecture notes cover the multimodal preference dataset

! What to watch out for

  • Setup requires Transformers and Torch to be installed
  • Images must be resized and converted before inference

FAQ

What is it built on?

The card names InternVL2.5 and mixed preference optimisation.

How is it loaded?

The documented path is AutoModel.from_pretrained with the Transformers library.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review