Zhejiang University Launches Qiushi Engine AI Research Tool with Thousand-Step Reasoning
Zhejiang University has unveiled a groundbreaking artificial intelligence system called “Qiushi Engine” (求是引擎) that can autonomously conduct complex scientific research through thousand-step long-range reasoning, marking a significant leap forward in AI-assisted scientific discovery. The system, developed by researchers at the university’s College of Information Science and Electronic Engineering, was announced on July 15, 2026, and represents China’s first AI system capable of independently organizing, executing, and iterating through multi-step scientific research workflows, according to China News Service.
What Makes Qiushi Engine Different
Unlike conventional AI research assistants that handle basic tasks such as literature review, code writing, and data processing, the Qiushi Engine is designed as an autonomous research system. Its core innovation lies in what its creators call “thousand-step long-range scientific reasoning” — the ability to break down complex research questions into thousands of discrete steps and execute them over hours or even tens of hours.
“‘Qiushi Engine’ is more akin to a new type of autonomous research tool,” said Yang Yihao, the system’s creator and a researcher at Zhejiang University, as reported by People’s Daily.
The system employs a multi-agent collaborative framework with five specialized modules: a Research Planning Agent that maps out research strategies, a Method Construction Agent that builds experimental and theoretical approaches, a Task Execution Agent that carries out specific research tasks, a Result Analysis Agent that evaluates outcomes, and a Risk Assessment Agent that identifies potential issues and checks for errors.
Proven in Real-World Experiments
In a validation experiment conducted on a real optical experiment platform, researchers provided the Qiushi Engine with only an open-ended research goal. The system then worked autonomously for over ten consecutive hours, completing literature review, theoretical analysis, experimental design, programming, data analysis, and result evaluation. After multiple rounds of failure and iterative correction, it produced multiple original research results — a workload that would typically take human researchers weeks to months.
“To enable the large model to effectively handle various complex situations in real scientific research scenarios, the Qiushi Engine is designed to interface with computing platforms, data analysis tools, experimental equipment, and specialized scientific software,” Yang Yihao explained, according to Guangming Daily.
Broader Implications for AI in Science
The launch of Qiushi Engine comes at a pivotal moment for AI-driven scientific discovery. Pan Yunhe, a Chinese Academy of Engineering academician and AI expert, framed the development within a broader global context.
“Global technological competition is shifting from ‘who has the more powerful large model’ to ‘who can use large models to solve complex problems in real-world scenarios,’ and scientific research is one of the most important directions in this regard,” Pan said, as reported by China News Service.
The system has already been deployed across more than ten research directions, including physics, optics, biomedicine, and mathematics. Notable achievements include proposing novel theoretical methods for decades-old problems in computational physics, building new theoretical frameworks in spectroscopy, and discovering new optical computing mechanisms.
What’s Next
Looking ahead, the research team plans to expand the Qiushi Engine into materials science, quantum science, and biomedical research. The system’s successful validation in real experimental settings suggests it could become a transformative tool in the scientific research infrastructure, potentially accelerating the pace of discovery across multiple disciplines.
As China continues to push the boundaries of AI application in science — exemplified by the 2026 World AI Conference held in Shanghai from July 17-20 — the Qiushi Engine represents a concrete step toward realizing the vision of AI as an autonomous research partner rather than merely a辅助 tool. Open questions remain regarding the reproducibility of its results, its performance across diverse scientific fields, and how it compares with similar systems being developed by institutions like Google DeepMind and Microsoft Research.
For now, Zhejiang University’s creation stands as a compelling demonstration of what happens when AI systems are designed not just to assist researchers, but to think alongside them.