{"ID":2848178,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.26742","arxiv_id":"2510.26742","title":"Running VLAs at Real-time Speed","abstract":"In this paper, we show how to run pi0-level multi-view VLA at 30Hz frame rate and at most 480Hz trajectory frequency using a single consumer GPU. This enables dynamic and real-time tasks that were previously believed to be unattainable by large VLA models. To achieve it, we introduce a bag of strategies to eliminate the overheads in model inference. The real-world experiment shows that the pi0 policy with our strategy achieves a 100% success rate in grasping a falling pen task. Based on the results, we further propose a full streaming inference framework for real-time robot control of VLA. Code is available at https://github.com/Dexmal/realtime-vla.","short_abstract":"In this paper, we show how to run pi0-level multi-view VLA at 30Hz frame rate and at most 480Hz trajectory frequency using a single consumer GPU. This enables dynamic and real-time tasks that were previously believed to be unattainable by large VLA models. To achieve it, we introduce a bag of strategies to eliminate th...","url_abs":"https://arxiv.org/abs/2510.26742","url_pdf":"https://arxiv.org/pdf/2510.26742v1","authors":"[\"Yunchao Ma\",\"Yizhuang Zhou\",\"Yunhuan Yang\",\"Tiancai Wang\",\"Haoqiang Fan\"]","published":"2025-10-30T17:38:14Z","proceeding":"cs.RO","tasks":"[\"cs.RO\"]","methods":"[]","has_code":false,"code_links":[{"ID":607599,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2848178,"paper_url":"https://arxiv.org/abs/2510.26742","paper_title":"Running VLAs at Real-time Speed","repo_url":"https://github.com/Dexmal/realtime-vla","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}