{"version":1,"type":"story","url":"https://digestai.news/story/kyutai-releases-reki-2-4b-vision-language-action-model-for-robot-manip","json":"https://digestai.news/story/kyutai-releases-reki-2-4b-vision-language-action-model-for-robot-manip.json","markdown":"https://digestai.news/story/kyutai-releases-reki-2-4b-vision-language-action-model-for-robot-manip.md","slug":"kyutai-releases-reki-2-4b-vision-language-action-model-for-robot-manip","headline":"Kyutai releases REKI 2.4B vision-language-action model for robot manipulation","summary":"Kyutai, together with Inria and École des Ponts, announced REKI, a 2.4 B parameter vision‑language‑action (VLA) model built on a 2.2 B Qwen3.5‑2B backbone plus a 0.2 B flow‑matching action expert. The model is pretrained solely on public data – a mix of 1.0 B video frames and 12.2 M vision‑language samples – and incorporates a cheap three‑frame memory that adds only 0.1 ms to inference. Benchmarked on a single NVIDIA H100, REKI processes an action chunk in 21.8 ms while using about 2 MiB of GPU memory per cached frame.\n\nOn out‑of‑distribution robot manipulation suites REKI scores 90.4 % on LIBERO‑Plus, the top result among publicly‑pretrained VLAs, and reduces pose‑oscillation failures on RoboTwin from 18.6 % to 4.1 %. Weights are released on Hugging Face under a CC BY‑NC‑SA 4.0 license, with six checkpoints covering the base model and finetunes for LIBERO, RoboTwin and RoboCasa variants. Deployment uses the LeRobot async‑inference gRPC protocol and can run on a UR5 arm after a short finetuning run.","keyPoints":["REKI combines a 2.2 B Qwen3.5‑2B VLM with a 0.2 B flow‑matching action expert, totalling 2.4 B parameters.","Achieves 90.4 % on LIBERO‑Plus and cuts RoboTwin pose oscillations from 18.6 % to 4.1 %.","Runs on one NVIDIA H100 at 21.8 ms per action chunk, with only 0.1 ms overhead for a 3‑frame memory cache."],"whyItMatters":"An open‑weight VLA with efficient memory enables more capable robot manipulation at modest compute, accelerating research and deployment in robotics.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Kyutai","Inria","École des Ponts","NVIDIA","LeRobot"],"models":["REKI","Qwen3.5-2B","Qwen-RobotManip","π0.5","Xiaomi-Robotics-1"],"people":[]},"firstPublishedAt":"2026-10-17T20:22:11Z","updatedAt":"2026-10-17T20:22:11Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"MarkTechPost","title":"Meet Kyutai REKI: The 2.4B VLA Model Bringing Efficient Memory to Robot Manipulation","url":"https://marktechpost.com/2026/10/17/meet-kyutai-reki-the-2-4b-vla-model-bringing-efficient-memory-to-robot-manipulation","publishedAt":"2026-10-17T20:22:11Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Kyutai releases REKI 2.4B vision-language-action model for robot manipulation\", 17 October 2026, https://digestai.news/story/kyutai-releases-reki-2-4b-vision-language-action-model-for-robot-manip","publisher":"Digest AI","title":"Kyutai releases REKI 2.4B vision-language-action model for robot manipulation","datePublished":"2026-10-17T20:22:11Z","url":"https://digestai.news/story/kyutai-releases-reki-2-4b-vision-language-action-model-for-robot-manip"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}