Skip to content

SmolVLM

A small open vision-language model

Visit official site

What SmolVLM is

The model is described as a small 2B parameter vision-language model that lowers inference cost and allows local deployment and customisation.

What you can do with it

  • Run multimodal AI in a small local setup
  • Analyse video when compute resources are limited
  • Evaluate on more benchmarks through VLMEvalKit

Application scenarios

  • Answer a travel question about a Bangkok landmark
  • Read a chart to locate drought regions
  • Pull dates out of an invoice image

What to watch out for

  • The named size is 2B parameters
  • Local use is described for browsers and edge devices
  • Loading uses Python and the Transformers library

Pros & cons

✓ What we like

  • Commercial use and custom deployment are allowed
  • VLMEvalKit integration supports more benchmarks

! What to watch out for

  • The described model size is 2B parameters
  • Local deployment is framed around browsers and edge devices

FAQ

How large is it?

The material describes a 2B parameter vision-language model.

What license is used?

The checkpoints, datasets, recipes and tools are released under Apache 2.0.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review