# Kyutai — REKI: Registering Knowledge in VLAs

- Company: Kyutai (kyutai.org)
- Announced: 2026-10-16
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://kyutai.org/reki
- Record: https://forck.live/items/23544-reki-registering-knowledge-in-vlas
- Subject: Moshi / Pocket TTS
- Open weights: yes
- Models affected: REKI

Kyutai introduced REKI, a 2.4B-parameter vision-language-action model (VLA) pretrained on public data. On LIBERO-Plus, RoboTwin Clean2Rand, and RoboCasa365, it is outperformed only by models about twice its size. The model uses a memory mechanism in its action expert that adds negligible cost and reduces oscillations between poses. Kyutai released the full code and all pretrained and finetuned model weights.

## Evidence

Verbatim from https://kyutai.org/reki:

> We introduce REKI, a 2.4B vision-language-action model (VLA) pretrained exclusively on public data that sets a new Pareto frontier between performance and model size: on LIBERO-Plus, RoboTwin Clean2Rand and RoboCasa365, it is outperformed only by models about twice its size.

---

Record: https://forck.live/items/23544-reki-registering-knowledge-in-vlas
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md