KorMedMCQA-V - X-Ray: leaderboard
Metric: Accuracy (%) on the questions carrying the 586 X-ray images (KorMedMCQA-V: image-based five-option questions from the Korean Medical Licensing Examination 2012-2023 (doctor track, R-type items excluded), each with one or more clinical images; zero-shot, closed-book, answer label parsed from JSON; open-source models averaged over three seeds); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 51 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 97 | #77 |
| 2 | Gemini 3 Flash | 96.2 | #93 |
| 3 | GPT-5.2 | 94.2 | #105 |
| 4 | GPT-5 | 93.6 | #91 |
| 5 | GPT-5 Mini (2025-08-07) | 89.1 | #165 |
| 6 | Qwen 3 VL 32B (Thinking) | 83.9 | #287 (Qwen 3 VL 32B) |
| 7 | Qwen 3 VL 30B A3B (Thinking) | 80.7 | #338 (Qwen 3 VL 30B A3B) |
| 8 | GLM-4.6V | 78.4 | #309 |
| 9 | Qwen 3 VL 32B Instruct | 75.9 | #276 |
| 10 | Qwen 3 VL 8B (Thinking) | 75.1 | |
| 11 | Qwen 3 VL 30B A3B Instruct | 72.2 | #365 |
| 12 | Ministral-3-14B-Reasoning-2512 | 71.2 | #519 |
| 13 | Ministral-3-8B-Reasoning-2512 | 64 | #694 |
| 14 | Qwen 3 VL 4B (Thinking) | 63.1 | #471 (Qwen 3 VL 4B) |
| 15 | Qwen 2.5 VL 32B Instruct | 62.3 | #443 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=kormedmcqa-v-x-ray · How It Works · Data refreshed daily, snapshot 2026-10-11.