Researchers at Stanford University have announced the debut of Biomni, an AI-powered multi-skilled biomedical research agent. Biomni is more than just a chatbot. This is a full-fledged “collaborative scientist” who can design and develop complex research workflows, said Jure Leskovec, the Alfred and Rebecca Lin Professor of Computer Science in the School of Engineering and senior author of the paper introducing Biomni in the journal. science.
“If you think of an agent as a carpenter, a carpenter without tools is just a carpenter who can talk,” Leskovec said, explaining how Biomni differs from typical generative AI chatbots. “At Biomni, we give carpenters the tools they need to build.”
Born to make an impact
Biomni was born from the idea that when using AI agents, scientists need to be able to explain their research problems in simple natural language. With this in mind, researchers designed Biomni to allow them to read the literature, develop a hypothesis, select datasets and tools, write code, interpret results, and suggest next-step experiments in a complete research workflow.
“Biomni can understand simple questions like, ‘Why do these patients respond differently to drugs?'” explains Kexin Huang, a former PhD student in Leskovec’s lab who recently completed his PhD and is now leading a startup company that will bring the technology to market. “Then there will be scientific research to dig deeper.”
Researchers chose biomedicine for its potential to improve the lives of everyday people. Scientific advances in biomedical research, from a fundamental understanding of life to new treatments for countless diseases, cannot be realized fast enough.
In a real-world example, a Biomni user uploaded over 450 files of continuous blood glucose monitoring, dietary intake, and physical activity data and asked a simple question: “Analyze this data to find interesting and plausible hypotheses.” In just 40 minutes, Biomni cleaned and integrated the data, generated visualizations, and identified patterns related to food intake and body temperature. Leskovec estimates that the task would have taken a human more than 60 hours to complete.
Biomni has another advantage over chatbots. It is to provide full citation and tracking of that work. Researchers claim that in its traceability, Biomni makes science more rigorous and reproducible.
If you compare an agent to a carpenter, a carpenter without tools is just a carpenter who can talk. Biomni provides a set of tools for carpenters to build.
Jure LeskovecProfessor Alfred Lin & Rebecca Lin
Speed of innovation
Biomni has specialized training in biomedical sciences. It incorporates extensive full-text published articles, code, and data stored in bioRxiv, a service that pre-publishes early versions of promising scientific findings to identify common software, tools, and databases used in biomedical research. The Biomni layer of 150 specialized biomedical tools, 105 software packages, and 59 databases spanning all 25 biomedical subdomains defined by bioRxiv, from genetics to neurology.
Biomni speeds up the process of scientific ideas and innovation. Leskovec explained that there is an inverse relationship between scientific information and the pace of discovery. As the amount of knowledge, data, and tools increased, innovation slowed down.
The reason for that slowdown is simple. Behind every breakthrough is years of research, and it all starts with a hypothesis. Just formulating a hypothesis requires a significant investment from scientists. Read the literature, ingest and homogenize datasets, write code, and look for unexplored patterns that form the basis of groundbreaking research. This process can take weeks or even months.
“The hurdle in biomedical science is not intelligence or ideas, but mechanics,” Leskovec stressed. “It’s this tedious work that slows down innovation. Biomni can do this in minutes.”
humans in the loop
Leskovec and Huang are quick to point out that while Biomni will not replace humans, it will allow them to focus on the values of scientists: ideation and judgment. Biomni can synthesize vast amounts of information and data very quickly and is adept at pattern recognition, but the choice to pursue a scientific trajectory requires human experience and reasoning.
“And it always will be,” Huang said. “This doesn’t mean machines will take over science, but rather that machines will become powerful new partners to augment human researchers. With Biomni, scientists will have a fast and energetic collaborator that allows them to focus on the important work of science.”
Biomni’s prototype is already in use in more than 10,000 laboratories in academia and industry, making it the most widely used AI collaborative scientist system in biomedicine.
“Biomni was my first research project that became widely available to real biologists,” Huang said. “It’s rewarding to be able to have such an impact on the way biologists work, and I’m excited to see where Biomni goes from here.”
reference: Huang K, Zhang S, Wang H et al.Autonomous biomedical research using artificial intelligence agents. science. 2026:eadz4351. doi: 10.1126/science.adz4351
This article has been reprinted from the following material: Note: Materials may be edited for length and content. Please contact the citation source for details. You can access our press release publishing policy here.
