Liang Wenfeng is the founder and CEO associated with DeepSeek, the Hangzhou-based Chinese AI company, and a co-founder and controlling figure of High-Flyer, an AI-driven quantitative-investment firm. He studied artificial intelligence at Zhejiang University, applied machine learning to quantitative trading, helped build High-Flyer’s computing and research capability, and launched DeepSeek in 2023. That sequence—not a sudden leap from hedge-fund management into startups—best explains DeepSeek’s emphasis on efficient computing, long-horizon research and openly released models.
Liang became internationally prominent after DeepSeek-R1 was released on January 20, 2025. Before then, he was better known in China’s quantitative-finance and AI circles than to the global technology audience.
Liang Wenfeng at a glance
| Fact | What is established |
|---|---|
| Name | Liang Wenfeng |
| Roles | Founder and CEO associated with DeepSeek; co-founder and controlling figure of High-Flyer |
| Education | Studied artificial intelligence at Zhejiang University, according to reporting by CNA |
| Location | DeepSeek is based in Hangzhou, China |
| DeepSeek founded | 2023, according to Forbes and Reuters |
| Breakthrough public date | DeepSeek-R1 release: January 20, 2025 |
| Public profile | Low-profile until DeepSeek-R1 drew global attention |
Published profiles differed between “39” and “40” because they appeared at different times and Liang’s birth date is not firmly established in the public record. Reuters described him as 39 in January 2025; it is safer not to state a current age.
From artificial intelligence to quantitative trading
Liang’s documented path starts with technical education rather than conventional finance. Reporting identifies him as an artificial-intelligence graduate of Zhejiang University. He then applied machine-learning methods to quantitative trading: using data, mathematical models and computing systems to make investment decisions rather than relying primarily on discretionary stock picking.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →That background matters because “AI investing” can mean two different things. Liang was not principally a venture capitalist choosing outside AI startups. His work combined quantitative finance, proprietary trading, internal machine-learning research and capital investment in computing infrastructure.
What High-Flyer is—and why it mattered
High-Flyer is a Chinese quantitative-investment and hedge-fund business founded by Liang and former university classmates. Its systems used algorithms, machine learning and large-scale computation to trade financial markets. A related entity often appearing in coverage is High-Flyer Quant; legal ownership and operating relationships among the entities should not be assumed to be identical.
Reported scale illustrates how the firm could finance research. CNA, citing official information, reported assets rising from about 1 billion yuan in 2016 to more than 10 billion yuan by 2019. Reuters later reported that the fund’s portfolio exceeded 100 billion yuan at the end of 2021. These are historical reported figures—not a current statement of assets under management.
High-Flyer also built AI capability for its own work. Reuters reported that it spent tens of millions of dollars on Nvidia hardware and researched overseas AI models before DeepSeek became famous. The firm therefore supplied more than money: it supplied researchers, data and engineering experience, access to expensive compute, and an organization able to pursue experiments over several years.
Recommended Free Tools
Why DeepSeek was a natural next step
The useful timeline is:
- Liang studied AI and moved into machine-learning-based quantitative finance.
- High-Flyer developed algorithmic trading systems and large-scale computing expertise.
- The firm invested in AI hardware and conducted model research.
- DeepSeek was established as a dedicated AI company in 2023.
- DeepSeek released successive model families, culminating in the R1 reasoning model’s January 2025 launch.
This makes DeepSeek less like a conventional venture-backed startup spun out of nowhere and more like a research laboratory built on an AI-intensive investment organization. High-Flyer could fund infrastructure and long-term work without depending immediately on consumer revenue or a large venture round.
Calling Liang an investor is therefore incomplete. His prior experience was in quantitative trading and AI infrastructure; those skills shaped the conditions in which a foundation-model company could be built.
DeepSeek’s model-development sequence
| Stage | What it shows |
|---|---|
| DeepSeek LLM | Early large-language-model work; official materials are available in the DeepSeek-LLM repository. |
| DeepSeek-V2 | A major efficiency-focused model generation documented in the V2 technical paper. |
| DeepSeek-V3 | A large mixture-of-experts model documented in the technical report and official repository. |
| DeepSeek-R1 | Reasoning-focused release announced January 20, 2025, with model and technical materials released under the MIT License: official notice. |
| DeepSeek-V4 Preview | Listed by DeepSeek’s official website as of August 18, 2026; “Preview” is part of the name: DeepSeek. |
What DeepSeek-R1 changed
DeepSeek’s January 20, 2025 announcement presented R1 as a reasoning model with performance comparable to OpenAI’s o1 on selected tasks. The company released model weights and technical materials under the MIT License. “Open source” is widely used to describe R1, but the precise openness depends on what is available: weights, code, data disclosures, training methods and licensing are separate questions.
R1 also challenged assumptions about the cost and hardware needed to produce competitive models. Reports often cite a training figure of roughly $5.6 million or less than $6 million. That should be read as a reported compute cost for a particular training run, not the total cost of building DeepSeek. It does not necessarily include salaries, data, networking, electricity, facilities, hardware purchases, failed experiments or earlier research.
Likewise, “cheaper than OpenAI” depends on the model version, token mix, caching, date and service. Capability comparisons depend on benchmark design, prompts, contamination controls and evaluator methodology.
How Liang’s background appears in DeepSeek’s technical strategy
DeepSeek-V3’s report describes a 671-billion-parameter mixture-of-experts model, with about 37 billion parameters activated for each token, pretrained on 14.8 trillion tokens. Its materials discuss Multi-head Latent Attention and other training and inference-efficiency techniques. The official repository provides deployment paths involving vLLM, SGLang, LMDeploy, TensorRT-LLM and LightLLM.
It is reasonable to interpret this emphasis on efficiency as consistent with Liang’s experience in quantitative modeling and infrastructure economics: compute efficiency, cost per token, distributed systems and owning or controlling technical capacity matter as much as headline parameter counts. That is an analysis of the organization’s trajectory, not proof that Liang personally designed every architectural choice. DeepSeek’s papers list large research teams.
How DeepSeek is financed and controlled
Public reporting links DeepSeek closely to High-Flyer and Liang. Reuters has described Liang as DeepSeek’s controlling shareholder and High-Flyer’s founder and controlling shareholder. A 2025 U.S. House Select Committee report also discussed a legally complex structure in which Liang retains effective control despite formal separation among entities.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteThe careful formulation is that DeepSeek is backed by, and closely linked to, High-Flyer within a structure controlled by Liang. “High-Flyer owns DeepSeek” is broader than the available evidence supports. Formal corporate ownership, financing relationships and practical control are not necessarily identical.
There is no dependable public basis here for stating a current valuation, personal net worth or current High-Flyer assets. Estimates describing Liang as a billionaire should be treated as media estimates, not verified balance-sheet facts.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why Liang remained so private
Liang had limited international visibility before R1. His profile rose sharply after the release and after he appeared at a symposium chaired by Chinese Premier Li Qiang on January 20, 2025.
Several factors may explain the contrast with highly public Western AI chief executives:
- DeepSeek presents itself as a research organization rather than a personality-led consumer brand.
- Its development was financed largely through a closely linked investment and infrastructure base rather than a heavily marketed venture process.
- Recruiting researchers and publishing technical work can matter more to the organization than constant founder publicity.
- China’s corporate and political environment differs from Silicon Valley’s media culture.
Those are plausible explanations, not a confirmed statement that privacy was deliberately chosen as a security strategy.
What Liang and DeepSeek say they are building
DeepSeek’s official website says the organization is dedicated to “exploring the essence of AGI.” A July 2026 Reuters report about a Yicai account said Liang prioritized long-term AGI development over maximizing short-term profit and that DeepSeek was likely to keep leading models open. Those comments came from reported investor-meeting material, not a published corporate policy or audited plan.
The distinction matters: prioritizing research, low prices or open releases does not make DeepSeek a nonprofit, and it does not prove that every future model will use the same license.
What remains difficult to verify
- Liang’s exact birth date, degree chronology and detailed early biography.
- The precise legal boundaries among DeepSeek, High-Flyer and High-Flyer Quant.
- Current assets, valuation, ownership percentages and Liang’s personal wealth.
- The complete cost of training and operating each model, beyond reported figures for particular runs.
- Claims about undisclosed chip inventories or prohibited hardware. Such allegations should not be presented as established facts without authoritative evidence.
Public model releases, technical reports and company statements are strongest for architecture and licensing. Independent evaluations remain necessary for broad claims that a model “beats” a named commercial competitor.
What his story means for understanding DeepSeek
Liang’s importance is not simply that a hedge-fund founder built an AI company. It is that an AI-intensive quantitative-investment firm became the financial and technical launchpad for a model lab. That origin helps explain DeepSeek’s focus on compute efficiency, infrastructure control, long-term research and open distribution.
For users, availability still has practical limits. Downloadable weights may require substantial GPU memory, quantization, distributed inference, compatible software and careful data-governance controls. Developers can use DeepSeek’s OpenAI-compatible API, while organizations with suitable hardware may self-host; the right choice depends on privacy, support, regional requirements and reliability—not token price alone. Current API prices should be checked on DeepSeek’s pricing page, because they can change.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
