GPT-6.1 Sol Cuts Localization Error in Live-Mouse Brain Surgery Tests
According to the disclosure, GPT-6.1 Sol was used in live-mouse brain surgery experiments and improved localization performance across three operations with surgeon feedback. [[fact:fact-bg2do4cb4]] By the third operation, time to identify key anatomical landmarks fell from 169 seconds to 86 seconds, while localization error declined from 0.6 mm to 0.3 mm. [[fact:fact-8i9gi81qe]] Exin Therapeutics co-founder and CEO Gabriel Ocana-Santero announced the results on X on November 5. [[fact:fact-zhbqtflgm]] The reported improvement came from accumulated experience and expert feedback from prior operations, rather than model retraining or parameter updates. [[fact:fact-urj591dbj]] Licensed veterinarians performed the incisions, craniotomy and suturing, supervised the entire surgery, and had to approve irreversible key steps. [[fact:fact-jur3pj2pf]] The research lead said three operations are a very limited sample and do not establish that AI can reliably master the skill. [[fact:fact-opogn0rvs]]