Mmastodon TechnologyRobotics first seen 19 h ago, last 8 h ago, peak #3
Kyutai releases REKI, a compact vision-language-action model for robots
Original: Kyutai has released REKI, a compact 2.4B vision-language-action model for robot manipulation. Pretrained on public data,
French AI lab Kyutai has released REKI, a 2.4-billion-parameter vision-language-action model for robot manipulation. Pretrained on public data, the compact model reportedly outperforms models twice its size on most benchmarks. All weights have been published on Hugging Face, letting researchers and developers download, run and fine-tune the model freely. The release is drawing attention in robotics circles as evidence that small, open models can rival much larger proprietary systems.
Why now: The open-weight release of a small model claiming to beat larger ones is notable news for the robotics and open-source AI communities.
Rank over time, top of the chart is #1. 9 snapshots from 19 h ago to 8 h ago.
Evidence
API: https://socialmediatrends-api.osmike.com/v1/trends/2686369