Medical Question Answering on MedXpertQA OOD (test)
69.2AccuracyGemini-3-flash
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini-3-flashFramework=Framework II, Inference Mode=Agent-based, Agent System=Ophiuchus2026.03 | 69.2 | |
| o3Model size category=Large Models2025.07 | 67.5 | |
| Gemini 2.5 ProModel size category=Large Models2025.07 | 58.9 | |
| Gemini-3-flashFramework=Framework II, Inference Mode=Direct Inference2026.03 | 54.4 | |
| Gemini 2.5 FlashModel size category=Large Models2025.07 | 47.4 | |
| Gemini-2.5-proFramework=Framework II, Inference Mode=Agent-based, Agent System=Ophiuchus2026.03 | 46.6 | |
| o3Framework=Framework II, Inference Mode=Agent-based, Agent System=Ophiuchus2026.03 | 44.1 | |
| GPT-5Framework=Framework II, Inference Mode=Direct Inference2026.03 | 40.4 | |
| Ophiuchus-7BFramework=Framework II, Inference Mode=Agent-based, Agent System=Ophiuchus2026.03 | 39.3 | |
| MeissaFramework=Framework II, Inference Mode=Agent-based, Agent System=Ophiuchus2026.03 | 36 | |
| Qwen3-VL-4BFramework=Framework II, Inference Mode=Direct Inference, Training Protocol=SFT2026.03 | 33.3 | |
| Gemma 3 27BModel size category=Small Models2025.07 | 29.8 | |
| MedGemma 4BModel size category=Small Models2025.07 | 24.4 | |
| Qwen3-VL-4BFramework=Framework II, Inference Mode=Agent-based, Agent System=Ophiuchus2026.03 | 23.9 | |
| Gemma 3 4BModel size category=Small Models2025.07 | 22.3 |