Skip to content

DriveVLM

Vision language models for autonomous driving

Visit official site
Category AI models

What DriveVLM is

The system is described as an autonomous driving approach using vision language models, with a dual variant that joins a conventional driving pipeline.

What you can do with it

  • Describe and analyse a driving scene
  • Produce layered planning predictions
  • Combine 3D perception for spatial reasoning

Getting started

  • Prepare an image sequence as input
  • Feed it into the model
  • Optionally add trajectory planning modules

Pros & cons

✓ What we like

  • nuScenes and SUP-AD datasets are used in experiments
  • DriveVLM-Dual was deployed on production vehicles

! What to watch out for

  • Image sequences must be prepared as input
  • 3D modules are optional add ons

FAQ

What is DriveVLM-Dual?

The material describes a hybrid system pairing DriveVLM with a traditional driving stack.

Which datasets are named?

The material names nuScenes and SUP-AD.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review