Rho-1
Reka's 19B omni-reasoning model that handles text, images, video and robot actions in one network, in research preview.
Overview
Reka announced Rho-1 on October 5, 2026 as a 19B-parameter model trained from scratch that reasons over and generates text, images and video, and can output robot actions from the same latent state. It unifies modalities as tokens in one context window instead of chaining specialist models, shows chain-of-thought before replies, and includes a distilled variant that cuts denoising from 99 steps to 8. Reka published no standard benchmark scores; access is by request.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- 19B parameters, trained from scratch across text, vision and actions
- Image generation, bounding-box grounding and video animation/editing in one conversation
- Real-time steerable world simulation with an 8-step distilled variant
- Robotics control predicting future observations and actions
- Native video output capped at 672x384
Use cases
- Multimodal reasoning over mixed text, image and video inputs
- Robotics research needing joint perception and action prediction
Pricing
Research preview; access by contacting contact@reka.ai. No pricing published.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 5 Oct 2026
- Record updated
- 15 Oct 2026
- Record last checked
- 15 Oct 2026