Skip to content

Molmo

An open multimodal model family that points at what it perceives

Visit official site

What Molmo is

The project is described as an open multimodal AI model family that learns to point at what it perceives to interact with physical and virtual worlds.

What you can do with it

  • Analyse and interpret multimodal research data
  • Build apps that interact with an environment
  • Improve product image and language processing

What it offers

  • Pointing-based perception of physical and virtual worlds
  • Small models described as outperforming models ten times larger
  • A robotics demo showing how AI can extend vision

Pros & cons

✓ What we like

  • The model family is described as open and actionable
  • Academic benchmarks and human evaluations are mentioned

! What to watch out for

  • Access steps include obtaining API permission
  • Robotics imagery is listed among the demonstrations

FAQ

What does the model learn?

The material says it learns to point at the content it perceives.

How do I get access?

The documented steps end with registering for API access.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review