Kyutai Unveils REKI, a 2.4B Model Enhancing Robot Memory
Kyutai, alongside Inria and École des Ponts, has released REKI, a 2.4 billion parameter vision-language-action model trained exclusively on public data. Featuring efficient memory for recent actions, REKI outperforms much larger competitors on out-of-distribution benchmarks. This release makes advanced robotic manipulation more accessible by providing open-weights models that match the performance of systems twice their size.
Source: MarkTechPost