Tether launches QVAC MedPsy, an offline medical AI for phones that beat Google MedGemma-27B on HealthBench Hard.
Tether has launched QVAC MedPsy, a medical AI model built to run offline on phones and consumer devices.
The company says the model topped Google MedGemma-27B on the HealthBench Hard clinical test suite, despite using far fewer parameters.
Tether Releases Offline Medical AI Model
Tether introduced QVAC MedPsy as a local-first medical health AI model. It is designed to work without a cloud link.
The company said the model can run fully on phones. This setup may help hospitals, clinics, and remote health teams with weak internet access.
QVAC MedPsy uses 1.7 billion parameters in one version. Tether also referred to a 4 billion parameter model with high processing efficiency.
8 billion humans deserve an intelligence that doesn't blink when the signal dies. 🧠
Introducing @QVAC Psy, our foundational models built on the mathematical stability of Psychohistory.With QVAC MedPsy, our local-first medical health AI model, we’ve proven that superior… pic.twitter.com/6ECt7kvk6Q
— Tether (@tether) May 7, 2026
The launch places Tether in the medical AI sector. The company is best known for USDT, the largest stablecoin by market value.
Tether said QVAC MedPsy was built on its QVAC AI system. It described QVAC Psy as a foundation model family using mathematical stability from “Psychohistory.”
The company said in a public statement, “8 billion humans deserve an intelligence that doesn’t blink when the signal dies.”
Model Tops Google MedGemma Test
Tether said QVAC MedPsy beat Google MedGemma-27B on HealthBench Hard. The test suite was scored by 262 doctors.
MedGemma-27B has 27 billion parameters. Tether said its 1.7 billion parameter model is nearly 16 times smaller.
The company framed the result as proof that design can matter more than size. It said, “Superior methodology beats raw parameter count.”
Tether ra mắt AI y tế siêu nhỏ, chạy được trên điện thoại và đánh bại mô hình lớn hơn 16 lần
Tether vừa phát hành QVAC MedPsy, một mô hình AI chuyên về y tế có thể chạy hoàn toàn trên điện thoại mà không cần kết nối đám mây. Mô hình 1,7 tỷ tham số của… pic.twitter.com/Kk7CKUd959
— ThuanCapital (@ThuanCapital) May 8, 2026
The benchmark result focuses on clinical reasoning tasks. These tests assess how models respond to complex medical questions.
The company also said its 4 billion parameter model uses fewer tokens than similar systems. It uses less than 3.2 times the tokens, according to Tether.
Lower token use can help reduce compute needs. It can also support faster replies on smaller devices.
Read Also:
Tether Q1 Earnings Top $1B With Reserves Climbing to New High
Local AI Could Support Remote Healthcare
Offline medical AI can be useful in places with limited internet. It can also reduce dependence on cloud systems.
Hospitals may use local systems when privacy rules are strict. Remote clinics may use them when network service is poor.
Running on consumer hardware could lower costs for some users. It may also support care teams during outages or field work.
However, medical AI remains a support tool. Experts often say doctors must keep final control over decisions.
The model’s test results do not replace clinical trials or regulatory review. Health systems usually need safety checks before using AI tools in care.
Tether said QVAC MedPsy is open source and sovereign. The company also called it a “tiny brain” for future local AI use.
The launch adds competition to medical AI development. It also shows growing demand for smaller models that work without cloud access.
For now, the main claim is clear. Tether says its offline medical AI runs on phones and topped Google MedGemma in a doctor-scored test.
