Unitree G1 Completes First Live Surgery on Living Subject

What actually happened in the operating room?

A research team at UC San Diego published results in Nature on July 8, 2026, describing two live laparoscopic cholecystectomies performed by a humanoid robot called Surgie, built on a Unitree G1 chassis. The 1.5-meter, 27-kilogram robot gripped standard commercial laparoscopic instruments from South Korean manufacturer LivsMed using a custom mounting bracket fitted to its hands. A surgeon controlled the system remotely via a stereo video headset and master tool manipulators. The teleoperation framework was designed by Lucas Zekai Liang, a PhD student in electrical and computer engineering at UCSD working in Michael Yip's laboratory.

The team ran two configurations. In the first, a single Surgie unit was teleoperated by a surgeon with a human assistant at the bedside. In the second, two Surgie robots worked side by side as a paired surgical team. Neither procedure required a fallback to conventional laparoscopy or open surgery, which is the standard benchmark for whether a robotic approach held up under real conditions.

Performance was mixed but meaningful for a first attempt. In dry-lab comparisons against the da Vinci research kit, Surgie achieved a weighted error of 4.53 versus 4.59 for the da Vinci system, indicating comparable precision on controlled tasks. In live surgery, tissue dissection and gallbladder detachment moved at a reasonable pace for an early prototype, but the procedure was interrupted repeatedly by pauses longer than three minutes. The cause was virtual remote center of motion drift, triggered by the animal's breathing and minor shifts in the robot's base position. System latency measured around 156 milliseconds.

Why does a general-purpose humanoid matter here, when dedicated surgical robots already exist?

The da Vinci surgical system, the dominant platform in robotic surgery, weighs 816 kilograms and costs between 1.8 and 2.5 million dollars. It also requires dedicated operating room modifications before it can be used. Surgie weighs 27 kilograms and can walk into any standard operating room without infrastructure changes. That physical difference is the core argument the UCSD team is making.

Approximately 67 percent of people worldwide lack access to basic surgical services, and a robot that costs a fraction of existing systems and needs no room modifications addresses a fundamentally different problem than dedicated surgical platforms.

Beyond surgery itself, the Unitree G1 platform is general-purpose. When not in the operating room, the same robot can move equipment, clean spaces, and handle other hospital support tasks. That dual-use potential changes the economics of deploying it. A hospital does not need to justify the cost of a single-purpose surgical robot; it can justify a robot that works across the facility and occasionally performs procedures.

The researchers draw a direct parallel to the first da Vinci cholecystectomy in 1997, which took hours to complete. The same procedure now takes roughly 30 minutes. The implication is that iteration time, not current performance, is the right metric to watch.

What are the real limitations, and what needs to happen next?

The study is candid about what does not work yet. The system has limited range of motion and insufficient force output for some surgical tasks. Calibration drift is a recurring problem, as the live surgery interruptions showed. There is no sterile solution for the robot's body, which is a non-trivial regulatory and safety barrier. Most importantly, the system has zero autonomy: every movement is surgeon-directed in real time. Surgie is a teleoperation tool, not an autonomous surgical robot.

Those limitations define the near-term research agenda. Force feedback, drift compensation tied to patient breathing cycles, and some degree of motion stabilization are the obvious engineering targets. Sterilization protocols will require collaboration with hospital infection control teams and likely regulatory guidance before any human trial could be considered. The path from animal subject to human patient involves FDA or equivalent oversight, institutional review, and probably years of additional validation.

For the robotics industry, the more immediate signal is that general-purpose humanoid platforms are now being seriously evaluated for high-stakes physical tasks outside of warehouses and factories. Unitree's G1 was not designed for surgery, yet a university team adapted it well enough to complete a live procedure. That says something about the flexibility of current humanoid hardware and about the appetite researchers have for pushing these platforms into new domains.