MedPsy AI model runs locally on phones, rivals larger Google model

Tether Data has released MedPsy, an open-source medical AI model designed for on-device execution, prioritizing privacy and efficiency. Available in 1.7B and 4B parameter sizes, the MedPsy-4B model demonstrates performance comparable to or exceeding Google's much larger MedGemma-27B on medical benchmarks. The development emphasizes superior training methodologies over sheer model size, utilizing high-quality synthetic data generated by Baichuan-M3-235B to achieve its results. AI

IMPACT Enables private, on-device medical AI applications, potentially reducing data privacy concerns and improving accessibility.

RANK_REASON Release of a specialized, open-source AI model with performance claims against a competitor. [lever_c_demoted from research: ic=1 ai=1.0]

Read on dev.to — LLM tag →

• Baichuan-M3-235B
• Hugging Face
• llama.cpp
• MedGemma-27B
• MedPsy
• QVAC Genesis II
• QVAC SDK
• Qwen3
• Tether Data

• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →