Skip to content

EAGLE

High-resolution multimodal models with mixed encoders

Visit official site

What EAGLE is

The family is described as vision-centric multimodal LLMs that combine vision encoders with different input resolutions to strengthen perception.

What you can do with it

  • Input resolutions above 1K
  • CLIP+X fusion of vision encoders
  • Pretrained models and fine-tuning data

Application scenarios

  • Read road signs and traffic signals for driving systems
  • Identify patterns in medical images
  • Handle image and text queries in customer service

Pros & cons

✓ What we like

  • Training and inference code are provided
  • Multiple input types including mixed modalities are listed

! What to watch out for

  • The named design mixes several vision encoders
  • The material singles out OCR and document understanding

FAQ

What resolution is supported?

The material states input resolutions above 1K are supported.

What is released with the models?

The material lists pretrained models, fine-tuning data and training and inference code.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review