Kyutai releases CADRE open video autoencoders for world models

Open causal video autoencoders cut world-model latency below 230 ms, aiding interactive robotics and simulation research.

CADRE is a family of 7 open causal diffusion video autoencoders for interactive world models. World models built on CADRE run at 20 fps on one H100 and react in under 230 ms.

Name

CADRE

Domain

creative

Read it at MarkTechPost →

Coverage: 1 sources, in order

1. 23 Oct 23:26 MarkTechPost Read the article →

← Front page