JMed48k (Text-Only): leaderboard
Metric: Accuracy (%) on the 9,905 text-only scored JMed48k-Eval items (2021-2025 exams), unweighted macro average over the 11 professions, official MHLW gold answers, single-shot Japanese prompts at temperature 0 with reasoning disabled where the model has a switch, JSON-constrained answers; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 95.5 |
| 2 | GPT-5 | 95.1 |
| 3 | GPT-5 Mini | 90.5 |
| 4 | Claude Sonnet 4 | 89.3 |
| 5 | Gemini 2.5 Flash (Non-reasoning) | 86.6 |
| 6 | Qwen 3.5 397B A17B (Non-reasoning) | 86.5 |
| 7 | Gemma 4 31B (IT) | 84.4 |
| 8 | DeepSeek R1 | 83.1 |
| 9 | Grok 4.20 | 82.4 |
| 10 | Qwen 3.5 27B (Non-reasoning) | 82 |
| 11 | Llama 4 Maverick | 77.7 |
| 12 | Qwen 3.5 9B (Non-reasoning) | 68.9 |
| 13 | Lingshu-32B | 65 |
| 14 | MedGemma-27B-IT | 60.6 |
| 15 | MedGemma-4B-IT | 34 |
Interactive version: theaggregate.ai/benchmark?slug=jmed48k-text-only · How It Works · Data refreshed daily, snapshot 2026-10-07.