KyutaiREKI
REKI 2B
The pretraining release combines public robot, simulation, human-hand and vision-language data. It is the starting point for the published task adaptations.
Research release
Register tokens carry recent visual context
REKI stores 32 tokens per frame across three frames one second apart. The action expert can use this recent history. The repository separates raw and moving-average weights; the documented fine-tuning path starts with raw weights.reki-2b: Kyutai model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | kyutai/reki-2b |
| Developers | Kyutai, Inria and École nationale des ponts et chaussées.reki-2b: Kyutai model card |
| Actual size | 2.4B parameters: a 2.2B vision-language component and 0.2B action expert.reki-2b: Kyutai model card |
| Ancestry and license | Qwen 3.5 2B; CC BY-NC-SA 4.0 weights and MIT code.reki-2b: Kyutai model card |
Related models
- REKI 2B LIBERO
Kyutai · REKI
A REKI action policy for the simulated Franka Panda robot in the four LIBERO evaluation suites.
Launch date not confirmedResearch releaseDownloadable - REKI 2B RoboCasa365
Kyutai · REKI
A REKI kitchen-task policy for a simulated Franka Panda arm on an Omron mobile base.
Launch date not confirmedResearch releaseDownloadable - REKI 2B RoboTwin Clean
Kyutai · REKI
A dual-arm AgileX ALOHA policy trained on the clean scenes of 50 RoboTwin tasks.
Launch date not confirmedResearch releaseDownloadable