Skip to content

InternVL2_5-1B

A multimodal model for images, video and text

Visit official site

What InternVL2_5-1B is

The model is described as part of the InternVL 2.5 series of multimodal large language models, with dynamic high-resolution training that improves work on multi-image and video data.

Getting started

  • Install torch and transformers
  • Load the model with AutoModel.from_pretrained
  • Prepare image and text inputs and preprocess the images

Pros & cons

✓ What we like

  • Random JPEG compression and loss reweighting are described for noisy images
  • A data filtering pipeline removes low-quality samples

! What to watch out for

  • The described setup relies on the ViT-MLP-LLM component chain
  • Inputs are prepared as images and text before inference

FAQ

Which LLMs are named?

InternLM 2.5 and Qwen 2.5 are listed as pretrained language models.

What does dynamic high-resolution training add?

The material states it improves handling of multi-image and video data.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review