The new family of models consists of two versions: MedPsy-1.7B, tailored for standard mobile devices, and MedPsy-4B, optimized for high-end phones and laptops. Despite their compact size, these models demonstrate superior results on clinical benchmarks compared to much larger counterparts. Specifically, the 1.7B model outperforms Google’s 27B MedGemma in realistic clinical assessments, while the 4B iteration maintains a competitive edge over models nearly seven times its size.
Tether launches QVAC MedPsy to bring medical AI to personal devices
Medical artificial intelligence no longer requires massive cloud infrastructure to function. Tether AI Research has introduced QVAC MedPsy, a suite of healthcare language models designed to run entirely on smartphones and laptops, ensuring patient data remains local while delivering reasoning capabilities that challenge significantly larger industry models.
By focusing on specialized medical training rather than broad parameter scaling, Tether achieved high accuracy with a 69 percent reduction in storage requirements. The models operate via the llama.cpp engine and the QVAC SDK, allowing for deployment in environments where internet connectivity is unavailable or privacy is a primary concern. As open-source software under the Apache 2.0 license, the platform requires no subscriptions or cloud-based accounts, mitigating risks associated with external data processing. This architecture provides a functional alternative for hospitals and remote clinics that need reliable diagnostic reasoning without the latency or security vulnerabilities of cloud-based infrastructure.
Comments (0)
No comments yet. Be the first!