RekaReka Rho
Rho-1
Rho-1 reads and writes text, images, video and robotic action representations in one model. Reka offers it as a research preview through collaboration.
Research release
What the release provides
The model was trained from scratch. Its understanding and generation streams share attention and state. Reka describes it as a proof of concept, with brittle editing, temporal limitations and native video resolution limited to 672 by 384.Rho-1 research preview
How to use it well
Practical guidance. Treat the preview as a research system. Inspect whether a generated sequence preserves objects and physical constraints before using it as simulation evidence.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Provider product |
| Access | Research preview by contacting Reka; no public weight or general API route is established by this announcement.Rho-1 research preview |