NeuroAI NEUROAINEUROAI.SITE
China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories
ESC

From Sogou to Baichuan: Wang Xiaochuan points China's AI at the clinic, not the chat box

After selling Sogou to Tencent, Wang Xiaochuan built Baichuan Intelligent and staked it on medical AI — a bet that the winner is the model doctors trust.

2026-11-06 · 972 words · NeuroAI
From Sogou to Baichuan: Wang Xiaochuan points China's AI at the clinic, not the chat box

Wang Xiaochuan stepped down as chief executive of Sogou (搜狗) on 15 October 2021, the day the search company he had built for two decades folded into Tencent (腾讯). Most founders that age would have taken the money and gone quiet. He announced a new company eighteen months later — and aimed it at hospital waiting rooms rather than app-store charts.

Key takeaways

  • Wang Xiaochuan (王小川) founded Baichuan Intelligent (百川智能) on 10 April 2023 with Ru Liyun (茹立云), the former Sogou COO; Wang had run Sogou as CEO since 2010 and negotiated its sale to Tencent.
  • Baichuan raised a US$50M angel round at launch and a US$300M Series A1 in October 2023 from Alibaba (阿里巴巴), Tencent and Xiaomi (小米), reaching unicorn status within six months of founding.
  • In July 2024 Baichuan closed a Series A of about RMB 5 billion (US$700M / HK$5.4 billion) at a roughly RMB 20 billion (US$2.8B / HK$21.7 billion) valuation, and on 31 August 2023 became the first model company founded that year to clear China's generative-AI service filing.
  • Wang's stated thesis is that medicine is the "crown jewel" of large models; Baichuan's assistant Baixiaoying (百小应) and an AI health adviser are the proving ground.
  • In its first year Baichuan shipped open models — Baichuan-7B/13B, Baichuan2-192K, Baichuan4 — at roughly 28-day intervals.

The engineer behind the brand

Wang is not a latecomer to artificial intelligence. A Tsinghua University (清华大学) computer-science undergraduate, he won an International Olympiad in Informatics gold as a teenager and spent years before the current wave applying machine learning to search, input methods and speech. That track record is why, when ChatGPT arrived, observers named him among the Chinese founders most likely to build a credible foundation-model company. He did not disappoint the timeline: he tried ChatGPT in January 2023, declared that "the era of strong artificial intelligence had arrived," and announced Baichuan that April.

The exit that was also a beginning

Sogou began inside Sohu (搜狐) in 2004. Wang joined Sohu in 2000, became Sogou's CEO when it was spun out in 2010, and turned an input-method tool into China's second-largest search engine. Tencent first bought into Sogou in 2013 with US$448M for a 36.5% stake. In 2020 Tencent offered US$9 per American depositary share; Sohu ultimately received about US$1.18B, and the merger closed in October 2021.

In his farewell letter Wang wrote that, over the next twenty years, he hoped "to contribute something to the life sciences and medicine, and to public health." Read today, that line reads less like a retirement note and more like a mission statement for the company he would soon start.

Why medicine, not another chatbot

Baichuan set health as a founding vision. At the 2024 World Artificial Intelligence Conference it demonstrated an AI health adviser that, like a general practitioner, asks follow-up questions — "any vomiting or fever? where exactly does it hurt?" — before offering a judgement. Per China National Radio, Baichuan's medical-enhanced model beat GPT-4 on the US Medical Licensing Exam (USMLE) in internal testing and on physician-led reviews by senior doctors from top hospitals.

In August 2024 Baichuan signed a strategic partnership with Beijing Children's Hospital (北京儿童医院), a national paediatric centre, to build a children's health model and a set of "smart paediatric" products for homes and grassroots clinics. Wang has called medicine "the crown jewel of large models," arguing that a model which can reason about symptoms does more to justify the technology than another writing assistant.

The funding arc

  • Angel: US$50M from Wang and friends at launch (April 2023).
  • Series A1: US$300M in October 2023, with Alibaba, Tencent and Xiaomi among the investors.
  • Series A: about RMB 5 billion (US$700M / HK$5.4 billion) in July 2024, at a roughly RMB 20 billion (US$2.8B / HK$21.7 billion) valuation.
  • Regulatory: Baichuan cleared the generative-AI service filing on 31 August 2023, the only company founded that year to do so at the time.

The pace mattered as much as the size. Baichuan published four open models and two closed ones in its first six months, a cadence the Beijing science bureau noted as the fastest among domestic model startups.

What the bet signals

Wang is making a quieter argument than his peers: that the durable value of a Chinese large model lies in a vertical where trust, not novelty, is the barrier to entry. If a model can sit beside a doctor and ask the right questions, the consumer-chat race becomes secondary. That reframes Baichuan from "another foundation-model contender" into a healthcare-infrastructure play — harder to copy, slower to monetise, and far less exposed to benchmark theatre. It also reflects a founder betting his second act on a problem he named publicly before he had a company to solve it.

Honest limitations

This profile draws on Baichuan's own announcements, government science-committee releases (Beijing and Ningbo), and reporting from China National Radio, China Daily and The Paper. The US$50M angel, US$300M Series A1 and RMB 5 billion July 2024 figures come from those outlets restating company statements, not audited filings; Baichuan is private and publishes no financials, so I have not independently verified revenue, headcount or deployment counts. The USMLE and physician-review claims about the medical model are company-reported test results, not peer-reviewed benchmarks, and I could not confirm them against a public leaderboard. Wang's "crown jewel" and "twenty years for medicine" remarks are quoted from company and media transcripts, not from a primary document I retrieved directly. Where a figure had only a single source, I dropped it rather than guess.

What readers can do now

  • Judge Baichuan by clinical deployments and doctor-trust signals, not by open-model leaderboard positions.
  • Watch whether the Beijing Children's Hospital partnership yields a shipped product rather than a demo.
  • Treat the RMB 5 billion / RMB 20 billion valuation as a company-stated number until a financing document confirms it.

Related coverage

Cite this article

NeuroAI (脑机网), "From Sogou to Baichuan: Wang Xiaochuan points China's AI at the clinic, not the chat box," 6 November 2026, https://neuroai.site/a/na-people-wang-xiaochuan-baichuan

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “Builders” → · Back to home · Markdown version

China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories