Reka Rho-1
#20 in Multimodal ModelsReka · v1 · since 5. Oktober 2026 · 2× · last seen Oct 07, 2026
25
Momentum
Rho-1 is a 19B-parameter omni model from Reka that processes and generates text, images, video and robot control actions within a single neural network. It was trained from scratch on 320 H100 GPUs over about three months. It is currently available only as a research preview; no public weights or public API have been released.
Momentum trend
13.07.11.10.
Features
| Key Benchmark (%) | Not verified in sources (no benchmark percentages found) |
| Multimodality | Input/output: text, image, video, robot actions, proprioception |
| Platform | Research preview, invitation/contact only; no public self-serve API |
| Release Date | October 5, 2026 (research preview) |