Chinese AI agent outperforms Anthropic’s Claude Code in autonomous research

A Chinese artificial intelligence system has achieved a significant milestone, outperforming leading international AI agents in autonomous scientific research. The Qiushi Engine, developed by a team at Zhejiang University, has claimed the top spot on the ResearchClawBench leaderboard, surpassing well-known systems such as Anthropic's Claude Code.

As of Tuesday, the Qiushi Engine held the overall leading position, with Open Science Desktop ranking second and Claude Code in third place. The ResearchClawBench is designed to evaluate the capabilities of AI agents in independently conducting scientific research. It measures their performance by comparing the results generated by the AI against reference papers.

This achievement highlights the rapid advancements in AI development within China, particularly in the complex field of scientific research. The Qiushi Engine's success suggests a growing capacity for AI systems to contribute to the scientific discovery process autonomously, potentially accelerating research and innovation.