Ant Group's LingBot model exceeds 2.2 million downloads; robots for pharmacies appear in Shanghai stores
Read more
Pandaily
pandaily.com

Ant Group's LingBot model exceeds 2.2 million downloads; robots for pharmacies appear in Shanghai stores

Ant Group's embodied artificial intelligence division announced that their open-source base robot models, LingBot, have been downloaded over 2.2 million times and garnered more than 33,000 stars on GitHub since their initial release nine months ago. This was reported by Leiphone during the Wuzhen World Internet Conference Summit, held from November 2 to 5, 2026.

The company introduced a comprehensive lineup of version 2.0 models designed to enhance the robots' capabilities in clearer vision, deeper reasoning, and faster action. This lineup includes the LingBot-Vision and LingBot-Depth 2.0 perception models, the LingBot-Video and LingBot-World 2.0 world interaction models, as well as the LingBot-VLA 2.0 and LingBot-VA 2.0 action models. The goal is to allow one model to be reused across different types of robots, tasks, and environments, eliminating the need to train a new 'brain' for every configuration.

According to the company, LingBot's work has been cited by the Physical Intelligence and Generalist AI projects. Stanford University's OpenWAM project adapted the LingBot-VA module, the Korean research project FastOPD utilized LingBot-VLA as a training model to train lighter robot policies, and EPFL applied LingBot-Map for three-dimensional reconstruction of coral reefs.

In terms of tools, LingBot-VLA and LingBot-VA were adapted for Hugging Face's LeRobot platform, LingBot-Vision was integrated into the main branch of the timm vision skeleton library, and LingBot-World is supported by vLLM and SGLang inference engines. NVIDIA included LingBot-World in its open-source world model inference library, FlashDreams. The first LingBot embodied models competition attracted 1,855 teams and 3,252 participants from 648 universities and companies across 13 countries during its online preliminary stage.

At the summit, the company demonstrated a robotic pharmacy in collaboration with Guoda Pharmacy, which replicates a real store with densely packed shelves. The robot operates in an 80-centimeter aisle, handling the entire cycle from order reception to item selection and handover. The company stated that this system is already operational in several Guoda Pharmacy stores in Shanghai, performing repetitive order fulfillment tasks instead of pharmacists, without any changes to product presentation or store layout.

In addition to pharmacies, the team is testing industrial loading and logistics sorting with partners, verifying the functionality of a single 'brain' on equipment from multiple robot manufacturers. Furthermore, a memorandum of understanding was signed with the Arab Federation for Digital Economy to study applications in retail, logistics, and hospitality in the Middle East.

Similar stories

Six humanoid AgiBot G2 robots operated in a Shanghai store for 24 hours, completing 1730 tasks
Read more
pandaily.com

Six humanoid AgiBot G2 robots operated in a Shanghai store for 24 hours, completing 1730 tasks

Six humanoid robots of the AgiBot G2 model completed a 24-hour shift in a convenience store located in the Xuhui district of Shanghai. This live-streamed event ran from 10:00 on November 2nd to 10:00 on November 3rd, and according to AgiBot, online viewership exceeded two million people.

Each robot performed a strictly defined function: one greeted customers at the entrance, two handled stocking, inventory replenishment, and shelf patrolling, one prepared coffee, one managed the checkout and hot food service, and the sixth was responsible for table cleaning. According to AgiBot's final count, the six robots completed 1730 tasks, which included 683 customer receptions, 369 deliveries and goods pickups, 147 coffees prepared, 126 breakfast orders, 277 inventory replenishment and shelf checking operations, and 128 instances of table cleaning.

Four out of the six robots passed the event certification, which assessed task completion speed, reaction time, ability to handle exceptions, and customer feedback. Invited testers attempted to create difficulties for the robots by mixing regular and diet cola and requesting only the latter, creating disorder on a full pack of chewing gum shelf, as well as standing in aisles or dropping boxes in the robots' paths.

Reports indicated that the robots confirmed beverages by combining voice requests with visual verification; they restocked shelves by visually counting items instead of relying on price tag placement; and they stopped to navigate around obstacles. Minor manipulations, such as hanging candy bags on hooks or picking up a soft cake without damaging it, remained among the most challenging tasks, and the robots occasionally made errors.

All six robots were coordinated using Genie Master, an embodied agent framework from AgiBot, which links task decisions, movement execution, and feedback through its Embodied Execution Protocol. The system analyzes successful and unsuccessful runs daily, stores useful experience in long-term memory, and filters corrected attempts for use in training data, while the robots practiced during periods of lowest activity. AgiBot emphasized that the test was conducted without human involvement in the loop, meaning no human demonstrations or on-site manual adjustments.

Ren Guanghui, Vice President of AgiBot's Embodied Research Center, told the media that the company would track metrics such as task success rate, first-attempt success, response time, total cycle time, qualified deliveries per unit of time, frequency of human intervention, recovery time, and system availability. However, the figures published upon completion only concerned the number of tasks completed, and AgiBot has not yet released data on success rates or the number of interventions.

This test launch marked the conclusion of AgiBot's AI WAVE series of releases, which began on October 22nd and integrated the company's datasets, simulation platform, models, and robotic hardware onto one live platform.

Ant Group releases LingGuang 2.0 assistant with persistent memory function for task management
Read more
technode.com

Ant Group releases LingGuang 2.0 assistant with persistent memory function for task management

Ant Group has introduced LingGuang 2.0, an updated version of its AI assistant. The new version is equipped with a persistent memory function, allowing it to continue managing tasks even after the user has ended the dialogue.

With appropriate permissions, the assistant can independently determine user needs, track changes, and perform subsequent actions over an extended period, instead of requiring a new command at every stage of work.

Examples of this function's use include checking forgotten recurring subscriptions in Alipay accounts, monitoring prices for flights and hotels after booking, and extracting tasks from school-related notifications in DingTalk groups.

The assistant integrates with services such as Alipay, DingTalk, Feishu, email, and the Xiaomi Mijia smart home platform. Furthermore, it supports external tools via command-line interfaces and MCP. The release of this version followed limited testing conducted earlier this month.

Sinomags, China's magnetic sensor production leader, plans IPO to develop robotic systems
Read more
technode.com

Sinomags, China's magnetic sensor production leader, plans IPO to develop robotic systems

Sinomags, a leader among Chinese manufacturers of magnetic sensors and holding the tenth position globally with a 0.6% market share, is preparing for an Initial Public Offering (IPO) in Hong Kong. These rankings are based on Frost & Sullivan data from 2025, according to information provided by the company after hearings.

The manufacturer from Anhui province presented a document to investors, emphasizing that its business integrates chip design, plate manufacturing, and sensor modules in both China and Germany. The company has achieved stronger positions in certain technologies and has recently become profitable.

The proposed listing will help finance further development, including expanding the production of precision sensors for areas such as robotics, as well as modernizing German production facilities acquired five years ago. For Sinomags, global expansion requires transforming accumulated competencies into a larger and more profitable business.

Magnetic sensors provide machines with information unavailable through simple visual inspection; they measure electric current, determine position, and track movement. For example, in a solar inverter, a current sensor monitors power flow, while in a robot joint, a magnetic encoder determines its position and trajectory.

These small components perform complex tasks, and their value depends on the ability to provide reliable measurements repeatedly under customer equipment operating conditions. Sinomags built its business around this requirement, starting operations in Ningbo in 2013 with a team of researchers in magnetic materials and electronics. Wang Jianguo gathered scientists, including Mao Xining and Bai Jianmin, to create a company whose commercial success depended on converting knowledge in magnetism into usable components.

Early focus areas for achieving production volume were solar energy and electric vehicles. By 2025, current sensors accounted for 82.6% of revenue. This segment gave Sinomags scale, a customer base, and manufacturing experience, closely linking its results to the needs of the energy and automotive markets.

The company's technological standing is stronger than its overall market share suggests. In the same 2025 rankings, Sinomags ranked second globally for Tunnel Magnetoresistance (TMR) sensors with a 4.7% share and fourth for current sensors with a 3.3% share.

TMR is one of the magnetoresistive technologies grouped under the general name xMR, which use changes in electrical resistance to detect magnetic fields. Sinomags operates across this family of technologies, and its position in TMR provides a foothold for expansion into a specific market segment.

A decisive step was the acquisition of Sensitec from Germany in 2021. Founded in 1999, Sensitec brought expertise in precision magnetic sensing and, critically, in plate manufacturing. This acquisition allowed Sinomags to link sensor chip design with their manufacturing and subsequent packaging into finished products for customers.

This is the essence of its Integrated Device Manufacturer (IDM) model. Owning most of the production chain gives the company greater control over how design transforms into a working component: magnetic materials, plate processes, chip characteristics, and module assembly all influence the final result.

Sensitec also provided the group with an established European business. The German company retained its name after the takeover and added Sinomags' current sensor portfolio to its European offering. This combines growing Chinese current sensor business with German capabilities in position and motion measurement.

Financial indicators show growth in the combined business. Revenue increased from 594.2 million yuan in 2023 to 845.3 million yuan in 2025. A loss of 66.6 million yuan turned into a profit of 9.9 million yuan in 2024, and then 52.6 million yuan in 2025. In the first five months of 2026, profit reached 47.9 million yuan on revenue of 418.7 million yuan.

Gross margin improved to 28% in the latest reporting period compared to 18% in 2023. However, the average selling price across the entire portfolio decreased from 7.16 yuan to 4.91 yuan, reflecting changes in products and sales structure. Sinomags earns more while selling a significantly higher volume of cheaper units on average.

This makes the structure of growth the central theme of its IPO. Volume growth strengthened the business, but entering more demanding sensing fields offers another path to increasing realized value.

Customer concentration creates pressure: the five largest clients account for over half of revenue, and the largest contributed 26.2% in the first five months of 2026. Thus, a relatively small group of buyers exerts a significant influence on the company's sales trajectory.

The offering documents also disclose a serious intellectual property dispute. A Chinese competitor in the magnetic sensor field challenged 25 of Sinomags' patents. Twenty were deemed invalid, two partially invalid, and three remained valid. Sinomags stated that it continues to challenge some decisions. The company asserts that the disputed patents relate to peripheral functions, while its most valuable technologies are protected as trade secrets.

These secrets include material preparation methods, temperature profiles, cutting parameters, and test data, as well as specialized equipment. These accumulated details determine whether a sensor can be produced stably at commercial scale. Revenue related to products containing the disputed patents dropped from 12% of the total in 2023 to 0.1% in the first five months of 2026. Commercial exposure has significantly decreased. Nevertheless, the dispute poses a pertinent question to potential shareholders: what part of Sinomags' advantage lies in enforceable intellectual property, and what part lies in manufacturing know-how that must remain in-house?

Future expansion will allow these insights to be applied in a different revenue structure. Robotics creates demand for precise position and motion measurement, and AI infrastructure opens opportunities in energy consumption monitoring and related components. For Sinomags, these markets expand from existing capabilities, although current sensors remain the highest revenue-generating business.

The planned use of IPO funds makes robotics ambitions tangible. Sinomags intends to build a scientific and production base in Xi'an with a maximum annual project capacity of 38 million high-precision magnetic sensors. Construction began in the first half of 2026, with operation expected in the first half of 2027.

Existing motion sensor module capacity is concentrated in Wettzell, Germany, at 6.5 million units annually as of May 2026. The new Chinese facility will bring production closer to domestic customers and expand the group's ability to serve robotics, automotive, and industrial applications.

There is also a telling detail behind the investments: Sinomags' current sensor lines in Benbu and Ningbo have spare capacity, but their equipment and processes cannot be easily transferred to motion sensor production. Transitioning to a new market requires additional production commitments.

The company also plans to use funds to upgrade its German plate manufacturing line from five-inch to eight-inch format. The acquisition that completed its production chain now requires new investment to improve efficiency and support further product development.

This gives the Hong Kong IPO a clear place in Sinomags' development. Research created the technical foundation, energy and automotive sector customers provided commercial scale, and Sensitec added manufacturing depth. Financing through the public market will support the next attempt to transform these elements into a broader precision sensing business.

The largest magnetic sensor manufacturer in China has secured its place in the industry. Its IPO calls on investors to fund the transition from this domestic position to greater global influence. What factories it builds, what products customers adopt, and what margins those products yield will determine how much its current 0.6% share grows.

Popular