Ant Group's Robbyant releases LingBot-VA 2.0, first native embodied foundation model
Original title:Ant Group’s Robbyant Unveils LingBot-VA 2.0: A Causal Video-Action Model Built Natively for Physical AI
AI summary
Robbyant, Ant Group's embodied intelligence team, released LingBot-VA 2.0, the first native embodied foundation model. The model uses a causal DiT architecture, with a video expert of about 13.0B parameters (about 1.9B activated), a training scale of about 15.3B parameters, and about 2.5B activated per token at inference. It introduces multi-chunk prediction (MCP) for 2.3x training acceleration and reduces inference latency to 142 ms/chunk through lookahead reasoning. On 50 tasks in RoboTwin 2.0, average success rates reached 93.8% and 93.4% on clean and randomized demonstration data, respectively.
Key Takeaways
- Robbyant,Ant Group's embodied intelligence team,released LingBot-VA 2.
- 0,the first native embodied foundation model.
- The model uses a causal DiT architecture,with a video expert of about 13.
- 0B parameters (about 1.
- 9B activated),a training scale of about 15.
Quick Facts
- Published
- Updated
- Source
- MarkTechPost(RSS)
- Category
- Model Releases
- AI score
- 74
具身智能领域第一个从零预训练的因果视频-动作模型,将世界状态和动作统一在一个隐空间,异步推理优化到 225Hz,做机器人的值得仔细研究。
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Ant Group's Robbyant releases LingBot-VA 2.0, first native embodied foundation model. July 11, 2026. https://new.nat6.net/en/news/cmrg3dsyt00g4iha7og0687tc/