China’s medical AI breaks ground as surgical robot wins EU approval, model tops benchmark

Medical AI from China has reached new milestones, with a teleoperated surgical robot gaining access to the European Union market and a clinical-grade model topping a major healthcare benchmark developed by OpenAI.
Shanghai MicroPort MedBot said its Toumai Remote robot, which allowed surgeons to remotely conduct laparoscopic surgeries, had received the “CE mark” from the European Union, a mandatory requirement to enter the market, according to its Hong Kong stock exchange filing on Monday.
The company said it was “the first remote surgical robot to obtain the CE mark”. The qualification indicated that a product met the requirements set out in EU rules allowing it to be “moved and marketed freely in the EU” regardless of where it was manufactured, according to the EU website.
According to MedBot, the Toumai robot consisted of three main components – a surgeon console, a patient cart and a vision cart. By integrating the latest 5G technology, it enabled remote procedures in urology, general surgery, thoracic surgery, and gynaecology.
Before the EU approval, the system was used in the UK, completing the country’s first robotic telesurgery, according to a BBC report in March. It enabled a London surgeon to remotely perform a prostate removal on a cancer patient in Gibraltar, some 2,400 kilometres (1,490 miles) away, according to the report.

Shanghai-based MedBot had delivered over 300 Toumai units across more than 60 markets worldwide, according to a post on its WeChat account on Monday.
It added that the EU approval marked a new era, with one of the most important Western markets “legally recognising teleoperation as a safe, effective and commercially viable medical approach”.
Overseas sales accounted for 73 per cent of MedBot’s 2025 revenue, up from 20 per cent in 2023, according to the company’s latest annual report.
Beyond hardware, Chinese healthcare-oriented AI has also advanced. Baichuan AI, a start-up backed by Alibaba Group Holding and Tencent Holdings, launched Baichuan-M4 on Monday. Alibaba owns the South China Morning Post.
The clinical-grade large language model (LLM) – designed to answer ordinary people’s questions on medical conditions and health preservation – topped HealthBench, an OpenAI benchmark for medical AI, outperforming general-purpose systems such as GPT-5.5 and Anthropic’s Opus 4.7, according to data provided by Baichuan.
The achievement aligns with a recent Oxford University research: while general-purpose LLMs achieved a 94.9 per cent diagnostic accuracy with complete medical charts from doctors, their performance plummeted to 34.5 per cent when chatting with ordinary people.
General LLMs often failed in real-world consultations because they relied on patients to provide a comprehensive medical history, with regular users without professional training struggling to describe their symptoms accurately, the study said.
Baichuan said its M4 model addressed this gap by transitioning from a passive question-answer tool into an active-inquiring agent. Instead of relying solely on baseline prompts, the model proactively interviewed users about the nature, triggers and history of their symptoms to narrow down potential conditions, it added.
Baichuan was eyeing an initial public offering by 2027, company founder and CEO Wang Xiaochuan told Chinese media in January.