KyutaiCADRE
CADRE 2x32x32
This CADRE autoencoder compresses video into a representation for interactive world models. One latent position covers 2 frames and a 32-by-32-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-2x32x32: Kyutai model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | kyutai/cadre-2x32x32 |
| Terms | CC BY-NC 4.0 weights, with DINOv3 terms for the frozen encoder.cadre-2x32x32: Kyutai model card |
| Training stage | 150,000-step pretraining checkpoint.cadre-2x32x32: Kyutai model card |
Related models
- CADRE 2x16x16 world models
Kyutai ยท CADRE
Four separately trained world models that predict video from context frames and actions in distinct environments.
Launch date not confirmedResearch releaseDownloadable