Ant Group's LingBot Robot Models Pass 2.2 Million Downloads as Pharmacy Robots Run in Shanghai Stores

Ant Group's embodied AI unit says its open-source LingBot robot foundation models have been downloaded more than 2.2 million times and have gathered over 33,000 GitHub stars in the nine months since their first release. Leiphone reported the figures from the 2026 World Internet Conference Wuzhen Summit, which runs November 2 to 5. The numbers are the company's own.

The 2.0 lineup

The company presented a full-stack 2.0 model lineup built around helping robots see more clearly, think more deeply and act faster. It spans three groups:

• Perception: LingBot-Vision and LingBot-Depth 2.0
• World models for interaction: LingBot-Video and LingBot-World 2.0
• Action models: LingBot-VLA 2.0 and LingBot-VA 2.0

The stated goal is to reuse one model across different robot bodies, tasks and settings instead of training a new brain for each. Parameter counts, benchmark scores and pricing were not part of the reported material.

Adoption and tooling

According to the company, LingBot work has been cited by Physical Intelligence and Generalist AI. It also names several downstream uses:

• Stanford's OpenWAM project reused and adapted the LingBot-VA module.
• A Korean research project called FastOPD used LingBot-VLA as a teacher model to train lighter robot policies.
• EPFL used LingBot-Map for 3D reconstruction of coral reefs.

On tooling, LingBot-VLA and LingBot-VA have been adapted to Hugging Face's LeRobot platform. LingBot-Vision has been merged into the main branch of the timm vision backbone library. LingBot-World is supported by both the vLLM and SGLang inference engines, and NVIDIA has integrated it into FlashDreams, its open-source world model inference library.

The first LingBot embodied model challenge drew 1,855 teams and 3,252 participants from 648 universities and companies in 13 countries in its online preliminary round.

Pharmacy deployment

Ant Group and Guoda Pharmacy showed a robotic pharmacy at the summit that recreates a real store with densely stocked shelves. A robot works in an aisle 80 centimeters wide and handles the full flow from receiving an order to picking and handing over items.

The company says the setup is already in regular operation at several Guoda Pharmacy stores in Shanghai, with no changes to product displays or store layout. It is meant to take repetitive picking off pharmacists' hands.

Next steps

Beyond pharmacies, the team is testing industrial loading and logistics sorting with partners, and validating one brain across several robot makers' hardware. It has also signed a memorandum of understanding with the Arab Federation for Digital Economy to explore retail, logistics and hospitality uses in the Middle East.

What this means

The download and star counts, and the citations, are company-reported, but the integrations are concrete: LeRobot, timm, vLLM, SGLang and NVIDIA's FlashDreams are existing open-source tools that now support LingBot components. That puts the models inside workflows robotics developers already use.

The Guoda Pharmacy deployment is the company's example of the models running in a live retail setting. The reported constraints are an 80-centimeter aisle and unchanged store layouts. The multi-manufacturer claim is still at the validation stage: the team says it is testing one brain across several robot makers' hardware, not that this is complete. Whether the 2.0 models deliver the reuse across robot bodies they target will depend on the results of those tests.