Physical AI / embodied world models
Meta V-JEPA 2
V-JEPA 2 is Meta’s video-based world model for understanding, predicting, and planning in physical environments. It is oriented toward robot reasoning and physical intuition rather than generating walkable 3D worlds for creators.
Overview
| Status | Research |
|---|---|
| Access | Paper / research |
| Released | 2025 |
| Inputs | video, image goals, robot observations |
| Outputs | latent predictions, planning signals, robot action support |
| Best for | physical reasoning, robot planning, video understanding, embodied AI research |
Why it matters
V-JEPA 2 helps define the embodied-AI side of the world model category: predicting outcomes, planning actions, and understanding physical dynamics.
Roamscape use
Tracked as a scientific world model reference for physical AI and embodied reasoning.
Strengths
- physical-world prediction
- planning orientation
- robotics relevance
- strong research framing
Limitations
- not a 3D world generator
- not a consumer creator tool
- outputs are not exportable worlds