Tether, the company behind the USDT stablecoin, has launched QVAC MedPsy, a medical AI model that achieves superior performance compared to larger rivals while being compact enough to run on consumer hardware. The 1.7 billion-parameter model outperforms Google's MedGemma-4B and even beats MedGemma-27B on the HealthBench Hard benchmark, which evaluates AI on realistic clinical conversations graded by 262 physicians. The 4 billion-parameter version generates responses with 909 tokens on average versus 2,953 for comparable systems—a 3.2x reduction. The models are shipped in quantized GGUF format (1.2 GB and 2.6 GB), enabling local deployment on smartphones, wearables, and hospital systems without cloud infrastructure. This addresses privacy concerns by keeping patient data on-device and avoiding HIPAA exposure. However, a recent Oxford study warns that LLMs still provide dangerous medical advice, suggesting AI should serve as a "secretary, not physician." For crypto wallet and key holders, this news highlights Tether's diversification beyond stablecoins into AI, but it has no direct security implications. Users should remain vigilant about phishing and social engineering attacks that may exploit the hype around Tether's AI releases. No wallet or key vulnerabilities are associated with this announcement.

Key facts

• Tether releases QVAC MedPsy medical AI model with 1.7B and 4B parameters.
• 1.7B model outperforms Google's MedGemma-27B on HealthBench Hard benchmark.
• Model uses 3.2x fewer tokens per response than comparable systems.
• Runs locally on consumer hardware, no cloud required.
• Oxford study warns LLMs still give dangerous medical advice.