HuatuoGPT-3: open-source medical LLMs via RL-only domain adaptation (OnePO)
CUHK-Shenzhen team's medical LLM series trained with one-stage policy optimization (no SFT); 27B variant scores 70.1 on HealthBench Total and 71.4 Professional, surpassing GPT-6 Astra.
Why it matters
Shows RL-only domain adaptation can beat the standard SFT+RL pipeline on a hard domain, with full training code, 20K RL tasks, and a grader model all released.
What you could build with it
A health-tech startup could reuse the OnePO pipeline to adapt a base model to its own clinical specialty with only ~20K verifiable examples, skipping the SFT stage; concrete result: a domain-expert assistant cheaper to train and closer to frontier quality.
Does it hold up?
Too early to judge: models and training code just released; no independent reproductions of the HealthBench scores yet.
Built with HuatuoGPT-3
- FreedomIntelligence/HuatuoGPT-3github · 16 ★ · OnePO training code, HuatuoGPT-3 9B/27B medical models, 20K medical RL tasks, rubric grader; Apache-2.0.
Learn more
First spotted on hf: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Mistral Large 4 (Le Chonk): 1.05T-param open-weight MoEMistral's new flagship: a 1.05T-parameter mixture-of-experts multimodal model with 49B active params and 1M-token…models · JEV 0.72ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71Mistral Forge: enterprise platform for training and continuously improving proprietary modelsMistral AI launched Forge (Sep 28) — a full-lifecycle model training platform (pre-training, SFT, DPO/ODPO, RL…models · JEV 0.69DeepSeek V4-Flash official API: public beta with upgraded agent capabilities, Responses API and Codex supportDeepSeek's official V4-Flash is now a public beta API with massively upgraded agent capabilities - benchmark scores…models · JEV 0.69
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.