Kyutai releases REKI, a 2.4B open robot manipulation model with cached memory
Overview
Kyutai, with Inria and École des Ponts, has released REKI, a 2.4B vision-language-action model for robot manipulation, pretrained only on public data, according to MarkTechPost.
REKI stores 32 cached register tokens across the last three frames, which the report says adds 0.1 ms to a 21.8 ms action chunk on one H100. Kyutai reports a 90.4% score on LIBERO-Plus, the top result in its comparison.
The weights are on Hugging Face under CC BY-NC-SA 4.0, which permits open use but restricts commercial use.
Written by AI from the articles below · updated Oct 17, 6:58 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- MarkTechPostNewsKyutai releases REKI, a 2.4B vision-language-action model with cached memory
AIKyutai, with Inria and École des Ponts, has released REKI, a 2.4B vision-language-action model for robot manipulation, pretrained only on public data. Its memory stores 32 cached register tokens across the last three frames, adding 0.1 ms to a 21.8 ms action chunk on one H100. REKI scores 90.4% on LIBERO-Plus, the top score in Kyutai's comparison, and its weights are on Hugging Face under CC BY-NC-SA 4.0.
Heat trend
Not enough continuous observations to show a trend yet.