HiMed-Trad: leaderboard
HiMed-Trad Bench from HiMed (Jiang et al., 2026): 6,010 Hindi multiple-choice questions on the seven Indian systems of medicine (Ayurveda, Yoga, Naturopathy, Unani, Siddha, Homoeopathy and Sowa-Rigpa), generated from OCR-digitized Hindi textbooks and filtered by experts. Scored as zero-shot multiple choice scored with a relaxed answer extractor, averaged over three runs; accuracy in percent, higher is better. The answer key is uneven (C 39%, B 34%).
Metric: Accuracy (%; 6,010 Hindi questions on Indian systems of medicine, zero-shot multiple choice, mean of three runs). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 62 | #741 |
| 2 | Qwen 2.5 7B Instruct | 53.3 | #2583 |
| 3 | Qwen 2.5 14B Instruct | 53.2 | #1745 |
| 4 | Llama 3.1 8B Instruct | 51 | #4277 |
| 5 | aya-23-8B | 46.6 | #6778 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=himed-trad · How It Works · Data refreshed daily, snapshot 2026-11-12.