Back to news
researchOpenAI2026-11-08

OpenAI GPT-6.1 Sol took part in live mouse brain surgery, halving localization error across three consecutive procedures

Exin Therapeutics disclosed that GPT-6.1 Sol located anatomical landmarks across three live mouse brain procedures, cutting identification time from 169 to 86 seconds and localization error from 0.6 to 0.3 millimeters.

On November 8, it was reported that according to earlier coverage and the latest disclosure from biotechnology company Exin Therapeutics, an OpenAI AI model, GPT-6.1 Sol, has been used in live mouse brain surgery experiments, improving localization performance across three consecutive procedures through surgeon feedback. By the third procedure, time spent identifying key anatomical landmarks fell from 169 seconds to 86 seconds and localization error dropped from 0.6 millimeters to 0.3 millimeters.

Exin Therapeutics co-founder and chief executive Gabriel Ocana-Santero published the results on a social platform on November 5. The team had GPT-6.1 Sol participate in anatomical landmark localization across three live mouse brain procedures, with professional surgeons correcting the positions identified by the AI after each procedure while the model saved images, measurements and correction records for reference in the next. The experiment mainly involved two key landmarks on the mouse skull surface, the bregma and the lambda.

According to the research team, after three rounds the manual correction actions required by GPT-6.1 Sol fell from four to two and localization error dropped 50 percent. Notably, this gain did not come from retraining the model or updating parameters, but from using experience accumulated in prior procedures and expert feedback to adjust its own judgment in subsequent tasks, a practical application of in-context learning.

GPT-6.1 Sol生物医疗小鼠脑手术上下文学习Exin Therapeutics