Tether releases locally runnable medical AI model QVAC MedPsy Odaily Planet Daily News: Tether AI Research Group has released a new generation of medical AI model QVAC MedPsy, which can run directly on low computing hardware such as smartphones and wearable devices, without relying on cloud servers. At the same time, it has surpassed multiple larger SOTA models in multiple medical benchmark tests. Official data shows that the 1.7 billion parameter version of QVAC MedPsy has an average score of 62.62 in 7 closed healthcare benchmark tests, which is 11.42 points higher than Google's MedGema-1.5-4B-it, although the model size is less than half of the latter. In real-world clinical tests such as HealthBench Hard, the model even surpassed MedGemma 27B with parameter sizes nearly 16 times larger. In addition, the 4 billion parameter version has an average score of 70.54, surpassing multiple large models with scales nearly 7 times larger in multiple medical inference evaluations. Tether stated that the model achieved "small model high performance" through post training medical reasoning optimization, reinforcement learning, and high-quality medical data training. Compared to traditional cloud based AI architectures, QVAC MedPsy significantly reduces inference costs. Its 4 billion parameter version generates an average of about 909 tokens, far lower than the 2953 tokens of similar systems, which can achieve lower latency and lower computing costs. The model also provides a quantified version of GGUF, suitable for local deployment on both mobile and edge devices. Paolo Ardoino stated that the core goal of this model is to improve its efficiency, rather than simply expanding the parameter scale, so that medical AI can run directly on local systems or terminal devices in hospitals, thereby avoiding sensitive medical data from being uploaded to the cloud.