Skip to content

Phi-3-vision-128k-instruct

A lightweight multimodal model with a 128K context

Visit official site

What Phi-3-vision-128k-instruct is

The model is described as a lightweight open multimodal model in the Phi-3 family for text and visual reasoning.

What you can do with it

  • Feed text and image inputs
  • Use chat format prompts
  • Work with a 128K token context

Application scenarios

  • Help students understand complex concepts
  • Analyze image and text data in business
  • Serve as a base model for generative AI research

Pros & cons

✓ What we like

  • Training used 512 H100-80G GPUs over 1.5 days
  • Supervised finetuning and preference optimization are described

! What to watch out for

  • Output is text generated for the input
  • Training data is described as 500 billion visual and text tokens

FAQ

How long is the context?

128K tokens.

What does the model output?

Text generated for the given input.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review