Stanford University researchers, including two Chinese PhD students, have released an open-source biomedical AI agent capable of designing and executing entire research workflows from plain-language instructions. The system, called Biomni, is already used by more than 10,000 scientists worldwide and can complete tasks that once took specialist teams weeks in just minutes, according to a study published this week in the journal Science.
Jure Leskovec, a Stanford computer science professor who supervised the work, said the agent was built to work alongside human scientists, not replace them. "We have over 10,000 scientists all over the world using the system for their everyday tasks," Leskovec said. The team released Biomni with a web interface so biologists could use it without writing code.
From plain-language queries to full research pipelines
The AI agent translates a plain-language request into a complete workflow. It searches databases, writes analysis code, identifies disease-causing genes, and generates step-by-step laboratory protocols. In one test, researchers gave Biomni hundreds of raw files from wearable devices and asked it to find biological patterns. The system cleaned the data, ran the analysis, and produced new hypotheses.
Kexin Huang, a PhD student in computer science at Stanford at the time of the work and an architect behind Biomni, said the team evaluated the system on more than 400 real-world research tasks. They compared its output against existing AI tools and human experts. In one case, Biomni completed a complex biological data analysis in 35 minutes that took a human expert three weeks. "It reached expert-level accuracy while cutting the time to just minutes," Huang said. Huang later co-founded Phylo, a San Francisco-based start-up that aims to make the AI system available to more researchers.
Dramatic time savings in real-world tests
Yuanhao Qu, a cancer biology PhD student at Stanford at the time and another main developer of Biomni, tested the system on a task he knew well: molecular cloning. The process of creating and replicating DNA constructs for experiments took Qu years to master and still consumed hours each time. When he gave the task to Biomni, "it did the whole thing from end to end," designing the experiment and producing a protocol that matched expert-level work in a fraction of the time.
"For me, Biomni is really changing the way biologists work," said Qu, who is also a Phylo co-founder and originally from Beijing. "Work that usually takes me hours now takes just minutes, so I can really spend my time on the science that actually needs a human."
Human oversight and current limitations
Leskovec stressed that the system is designed as a collaborator. Humans still ask the questions, judge the results, and decide which directions to pursue. "Biomni is a powerful tool, not a decision maker," he added.
The researchers acknowledged that Biomni has only been tested on part of the biomedical field. It still struggles with tasks requiring deeper scientific judgment, original experimental ideas, and complex reasoning. Future versions could become broader and more self-improving by learning from experience, incorporating more types of scientific data, and continuously adding new research tools and knowledge.
Why this matters for science and research professionals
Biomni demonstrates that routine, time-intensive research tasks can be automated with expert-level accuracy, freeing scientists to focus on creative problem-solving and experimental design. The open-source release means research groups and institutions can adopt the tool immediately, without licensing barriers. The key takeaway is not that AI will replace researchers, but that it can handle the repetitive parts of the workflow, provided human experts remain in the loop to verify outputs and set direction.
Your membership also unlocks:
