Liang Wenfeng on Innovation

Liang Wenfeng's DeepSeek deepSeek's V4 model arrives mid-February 2026 with Engram memory architecture, targeting Claude and GPT supremacy in code generation.

DeepSeek's leadership articulated a strategic priority on advancing toward artificial general intelligence as a long-term objective, weighing that goal ahead of near-term profit maximization, a framing that accompanied the company's open-source model-release cadence.

Through High-Flyer, Liang built large-scale computing infrastructure originally for quantitative finance that was later repurposed to support AI research, leading to DeepSeek's founding in July 2023.

In July 2026, Bloomberg News reported Liang had become the world's wealthiest founder of an AI-model company, with a net worth of $36 billion per the Bloomberg Billionaires Index; in 2025 he was named to Time's list of 100 Most Influential People in AI and Fortune's Most Powerful People in Business.

Liang Wenfeng's DeepSeek initializing from the V3 model and sharing the V3 architecture, DeepSeek was planning to release its first chatbot application, the DeepSeek-R1 .

DeepSeek's January 2025 release of the DeepSeek-R1 model drew international attention for demonstrating performance comparable to OpenAI's GPT-4 and o1 while reportedly using significantly lower training costs (the company claims about $6 million to train its V3 model, versus OpenAI's reported ~$100 million for GPT-4) and roughly one-tenth the compute of Meta's comparable Llama 3.1 model.

Liang Wenfeng's DeepSeek 2023, Liang formally established DeepSeek, positioning it as a pure research laboratory with artificial general intelligence (AGI) as its .

Liang Wenfeng's DeepSeek deepSeek's January 2025 release of R1 disrupted prevailing assumptions about the cost of frontier capability, and its April 2026 V4 release .