Apex Intelligence raises nearly US$50M angel to build self-evolving foundation models

Apex Intelligence said it completed angel and angel-plus funding rounds totaling nearly US$50 million. The angel round was co-led by IDG Capital, LinkX Capital, and XtalPi; the angel-plus round was co-led by Zhongguancun Science City Fund, SCGC, and Shanghai Engine Fund. Founded June 2026 by Tsinghua AI assistant professor Yongchao Chen, the company says it is building recursively self-improving foundation models aimed at scientific research and discovery.

FUTURE desk — same-day primary that a brand-new Beijing lab raised nearly US$50M at angel stage to chase recursively self-improving models for science, a concrete money signal on the RSI race.

The angel round, still company: co-led by IDG Capital, LinkX Capital, and XtalPi, with participation from Decent Capital, SEE Fund, Monad Ventures, and Winsoul Capital. None was printed.

The angel-plus round, still company: co-led by Zhongguancun Science City Fund, SCGC, and Shanghai Engine Fund. The wire does not split the nearly-US$50 million between the two rounds.

Company thesis, as the same wire has it: founded June 2026 in Beijing; building the next generation of self-evolving foundation models. Self-evolution, the company says, means Recursive Self-Improvement, or RSI — models that improve themselves. The focus is scientific research and discovery. The company says it wants to move AI from an assistive tool to a co-researcher, and from accelerating what is known to discovering what is unknown. A foundation model is a general model you later adapt to many jobs.

Founder and team, still company: Yongchao Chen, assistant professor at Tsinghua University’s School of Artificial Intelligence. The same wire says his undergraduate and graduate path spans the University of Science and Technology of China (USTC) and a joint Harvard–MIT training program, and that he has conducted research at Google DeepMind, Microsoft, and the MIT–IBM Watson AI Lab. The core team, the company says, comes from Tsinghua, Peking University, Harvard, and MIT. Extra career color and the founder’s quotes stay in Sources.

Talent and collaboration plan, company: Apex Intelligence says it will visit leading North American universities — Harvard University, MIT, Yale University, and Boston University — from September 17 to September 23 to meet students, researchers, and academic research groups for talent recruitment and research collaboration. This is not a claim those visits have already happened, and it is not a signed university partnership.

Research claims on the same wire, company assertions only — not independently confirmed breakthroughs: in “AI for AI,” the company says its self-evolving system has autonomously produced research of a standard suitable for acceptance at leading AI conferences, and that it has set new records or approached the best publicly reported results on named benches including SimpleTES, NanoChat Autoresearch, GPUMode TriMul, and MLS-Bench. In “AI for Math,” the company says the system produced a complete proof of the majorization conjecture, which the wire says had remained open for more than three decades, plus other results in optimization theory and geometric topology.

Use of funds and long-term goal, still company: continue investing in foundation models and compute, advance training and iteration of recursively self-improving foundation models, build a high-quality research-trajectory data infrastructure, and attract top talent. The about box calls Apex Intelligence a Beijing, China-based self-evolving foundation model company focused on next-generation industrial intelligence through recursive self-improvement, and points to https://apexin.ai/. Named possible application fields on the wire: chip design and simulation, molecular R&D, and quantitative strategy iteration. The “100,000 top-tier AI scientists” line is the company’s long-term vision.

Plain English for the rest of the card: angel = very early venture money, usually before a classic seed or a later lettered growth round. Angel-plus = a second early check after the first angel round, still before a priced growth round. RSI = recursive self-improvement, models that try to improve themselves. Self-evolving = the company’s word for that RSI loop. Foundation model = a general model you later adapt to many jobs. Hypothesis–experiment–validate = the company’s described loop: guess, test, check, then try again, rather than only copying human preference data. Tsinghua = Tsinghua University in Beijing. USTC = University of Science and Technology of China. Majorization conjecture = a math claim the company says its system proved; treat that as a company assertion, not a journal result this desk checked.