Openai's GPT-6.1 improves surgical precision in live mouse brain experiments through contextual learning

Exin Therapeutics disclosed that OpenAI's GPT-6.1 Sol was utilized in live mouse brain surgery trials, demonstrating significant performance gains across three consecutive operations. According to the findings released by Gabriel Ocana-Santero, CEO and co-founder of Exin Therapeutics, the model's time to identify key anatomical landmarks dropped from 169 seconds to 86 seconds, and its positional error decreased from 0.6 mm to 0.3 mm. This improvement was achieved not through retraining, but by leveraging accumulated experience and expert feedback from professional surgeons who corrected the AI's positioning after each session. The team also noted that the required manual corrections fell from four to two instances. These tests involved identifying Bregma and Lambda points on the mouse skull for stereotactic procedures. Previously, the research group had tested Claude Sonnet 5, GPT-6.1 Sol, and xAI's Grok 4.7 on gene therapy selection for epileptic mice. While all three models proposed different treatment mechanisms—Claude chose the entorhinal cortex, GPT-6.1 Sol selected the hippocampus, and Grok chose the somatic sensory cortex. The experiments, which are in an early exploratory phase, confirm AI's potential for direct device operation, though licensed veterinarians maintain oversight for critical steps.