V-JEPA 2
$ 0
Video Joint Embedding Predictive Architecture; predicts future video representations for robot action planning from passive video
Out of stock
Brain Score
3
Specifications and details:
| Nationality | US |
|---|---|
| Website | https://ai.meta.com |
| Model type | Foundation Model |
| Manufacturer | Meta AI |
| Release date | 2025 |
Description
V-JEPA 2 is a foundation model that helps robots learn by observing how the world changes over time. Instead of relying on explicit action labels, it predicts how scenes will evolve in the near future. Consequently, the system develops a deeper understanding of motion, cause, and effect. Moreover, this predictive ability allows robots to anticipate outcomes, improving timing, coordination, and task execution across diverse scenarios.
2026 Humanoid Robot Market Report
160 pages of exclusive insight from global robotics experts – uncover funding trends, technology challenges, leading manufacturers, supply chain shifts, and surveys and forecasts on future humanoid applications.
Featuring insights from
Aaron Saunders, Former CTO of
Boston Dynamics,
now Google DeepMind
2026 Humanoid Robot Market Report
160 pages of exclusive insight from global robotics experts – uncover funding trends, technology challenges, leading manufacturers, supply chain shifts, and surveys and forecasts on future humanoid applications.
In addition, V-JEPA 2 learns from passive video without requiring detailed annotations. Therefore, it scales efficiently across massive datasets, capturing a wide range of real-world behaviors. This learning strategy enables robots to adapt to unfamiliar situations with greater confidence. Because it emphasizes prediction over imitation alone, the model supports flexible decision-making and proactive control. As a result, advances humanoid systems toward intuitive, forward-thinking intelligence.
Contact Humanoid.guide
Website: https://ai.meta.com
