Updated

Rho-1 (research preview)

Reka released a research preview of Rho-1, a 19B omni-reasoning model trained from scratch. The company says it can understand and generate text, images, and video, reason across them, and take actions within one model.

By No Code MBA · Facts checked

Provider
Reka
Announcement date
Status at last check
Limited access or preview
Model family
Rho
Capabilities
Text, Image, Video

What changed

Rho-1 is a 19B model trained from scratch that takes in and generates text, images and video, and also outputs robotic actions, within a single model and context window. Reka says it trained on 320 H100 GPUs for three months, that the base model generates video at 0.79x real time, and that a distilled variant cuts denoising from 99 to 8 steps, returning a 5.3 second clip in about one second.

Access and availability

Released as a research preview; Reka's post invites collaboration through contact@reka.ai and does not describe a public API, playground or downloadable weights.

Official access documentation

What you can build

If you can access the preview, test a task that combines visual understanding, generation, and follow-up editing, then review each output for consistency and errors.

Limitations

Reka lists long-horizon drift, object grounding that works on images but not yet across video, brittle targeted editing, and native video capped at 672x384. Context length and license are not published.

Sources and release history