Tether's AI research team has unveiled the QVAC MedPsy medical language model, designed to operate on low-power devices like smartphones and wearables. This model aims to deliver performance comparable to larger models while ensuring full localization and privacy protection. The 1.7 billion parameter model achieved an average score of 62.62 across seven medical benchmarks, outperforming Google's MedGemma-1.5-4B-it by 11.42 points, despite having fewer parameters. Tether CEO Paolo Ardoino stated that this initiative seeks to transform the application of medical AI by enabling local execution of medical reasoning on-site, such as within hospital systems or mobile devices, without relying on cloud processing of sensitive information.