Meet the team that turns your ideas into work.

Discover EightBots
EIGHTBOTS
Contact salesSign up
KyutaiCADRE

CADRE 4x24x24

This CADRE autoencoder compresses video into a representation for interactive world models. One latent position covers 4 frames and a 24-by-24-pixel region, with 64 channels.

Research release

The codec reconstructs video from compressed features

The published setup uses 20-frame-per-second video at 288 by 512 pixels. The causal flow-matching decoder reconstructs frames from the latent representation. The model is an autoencoder, not a complete action-conditioned simulator.cadre-4x24x24: Kyutai model card

Specifications & access

CADRE 4x24x24 specifications
DetailValue and conditions
StatusResearch release
AccessDownloadable
Model identifierskyutai/cadre-4x24x24
TermsCC BY-NC 4.0 weights, with DINOv3 terms for the frozen encoder.cadre-4x24x24: Kyutai model card
Training stage150,000-step pretraining checkpoint.cadre-4x24x24: Kyutai model card

Related models

  • Four separately trained world models that predict video from context frames and actions in distinct environments.

    Launch date not confirmedResearch releaseDownloadable