Reportedly: Ant Lingbo showcases fully open-source robot general brain at Wuzhen Summit
Reported by Leiphone · not yet confirmed by the company or a second independent outlet. We update this page when it is.
Ant Group's embodied AI unit Ant Lingbo presented its fully open-source robot general brain at the 2026 World Internet Conference Wuzhen Summit. LingBot series models have surpassed 2.2 million downloads and 33,000 GitHub stars in nine months. The models are integrated into LeRobot, timm, vLLM, SGLang, and NVIDIA FlashDreams, Leiphone reported.
China context
- Original name
- 蚂蚁灵波
- Outside China
- Open weights · leiphone.com
- Claims
- Company-reported; not yet independently evaluated
- For builders
- Developers outside China can integrate LingBot models via LeRobot, timm, vLLM, SGLang, and NVIDIA FlashDreams, reducing the need to train robot models from scratch.
- For investors
- The open-source traction (2.2M downloads, 33K stars) and commercial pharmacy deployment indicate a viable path for embodied AI foundation models, but revenue model and international commercial adoption remain unverified.
Translated from Chinese. Quotes and facts link to the original sources.
Ant Lingbo, the embodied AI company under Ant Group, showcased its fully open-source robot general brain at the 2026 World Internet Conference Wuzhen Summit (November 2-5). The company has open-sourced core models covering spatial intelligence, dexterous manipulation, and environmental feedback. LingBot series models have achieved over 2.2 million downloads and 33,000 GitHub stars since release nine months ago. Global research institutions including Stanford University, EPFL, and Korean researchers have adopted or adapted LingBot models. The models are integrated into major developer toolchains: LingBot-VLA and LingBot-VA are adapted for LeRobot, LingBot-Vision is merged into timm, LingBot-World is supported by vLLM and SGLang, and NVIDIA integrated LingBot-World into FlashDreams. Ant Lingbo also demonstrated a robot smart pharmacy with Guoda Drugstore, where robots autonomously complete order picking and delivery in 80cm-wide aisles; the solution is in regular operation at multiple Shanghai stores. The first Ant Lingbo Embodied Large Model Challenge attracted 1,855 teams and 3,252 participants from 13 countries and 648 universities/companies.
The LingBot model matrix includes LingBot-VLA (vision-language-action), LingBot-VA, LingBot-Vision, LingBot-Map, and LingBot-World. LingBot-World is the first world model to receive support from both vLLM and SGLang open-source inference engines. The models are designed for cross-embodiment reuse, allowing the same intelligence to be applied across different robot configurations, tasks, and scenarios.
Robot developers outside China can now use Ant Lingbo's open-source models to avoid training robot brains from scratch for each robot configuration and task, reducing development cost and time. The integration into LeRobot, timm, vLLM, SGLang, and NVIDIA FlashDreams lowers the barrier for global developers to adopt these Chinese models in their own robotics stacks.
Ant Lingbo's open-source strategy positions its robot general brain as a foundational layer for embodied AI, potentially driving adoption in enterprise scenarios such as pharmacies, industrial loading/unloading, and logistics sorting. The partnership with Guoda Drugstore demonstrates a working commercial deployment in Shanghai, and collaborations with partners like Leju Robotics and Titanium Tiger explore 'one brain, multiple robots'.
Next observable signals include whether LingBot models are adopted by more non-Chinese robotics companies beyond the cited research projects, and whether the partnership with the Arab Digital Economy Alliance leads to commercial deployments in Middle East retail, logistics, or hospitality. The outcome of the Ant Lingbo Embodied Large Model Challenge finals may also indicate developer traction.