Skip to content

SpatialVLM

Reason about spatial relations in images

Visit official site

What SpatialVLM is

The model is described as a vision-language model that gives spatial reasoning ability through large-scale synthetic training data.

What you can do with it

  • Reason about qualitative spatial relations
  • Estimate distance and size quantitatively
  • Run chained multi-step spatial reasoning

Application scenarios

  • Judge which of two objects is closer to the camera
  • Estimate the horizontal distance between two objects
  • Decide whether a table forms an equilateral triangle

Pros & cons

✓ What we like

  • Chained multi-step spatial reasoning is supported
  • Rewards can be provided for robot control

! What to watch out for

  • Trained on large-scale synthetic data
  • Reasoning centres on spatial relations in images

FAQ

Who developed it?

The material names Google DeepMind.

What can it estimate?

The material lists distances and sizes quantitatively.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review