Wu says he has learned that Tether’s AI research team today launched the QVAC MedPsy medical language model. The model series is designed to run directly on low-compute devices such as smartphones and wearable devices, delivering performance comparable to large models while enabling full local deployment and privacy protection. Data shows that its 1.7B-parameter model scored an average of 62.62 across seven medical benchmark tests, outperforming Google MedGemma-1.5-4B-it—by 11.42 points, despite the latter having more than twice the number of parameters. Tether CEO Paolo Ardoino said this initiative is intended to change the application limitations of medical AI, so that medical reasoning can run locally where the data resides (such as hospital systems or mobile devices), without relying on cloud processing of sensitive information.