Med-CoReasoner - MMLU-ProX-Health: leaderboard
Baseline results from the Med-CoReasoner paper (Gao et al., 2026) on the health category of MMLU-ProX (687 multiple-choice items per language), averaged over Chinese, Japanese, Korean, German, French, Spanish, Italian and Swahili. Exact-match accuracy (%) with sampling temperature 0.7 and low reasoning effort for reasoning models; higher is better.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 (Low) | 74.53 | #172 (GPT-5.2) |
| 2 | GPT-5.1 (Low) | 71.39 | #225 (GPT-5.1) |
| 3 | GPT-4o | 67.94 | #741 |
| 4 | DeepSeek V3.2 | 65.15 | #403 |
| 5 | Claude 3.5 Haiku | 56.21 | #1563 |
| 6 | Qwen 2.5 72B Instruct | 55.45 | #1093 |
| 7 | Llama 3.1 70B Instruct | 54.07 | #1541 |
| 8 | Qwen 2.5 32B Instruct | 53.15 | #1216 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=med-coreasoner-mmlu-prox-health · How It Works · Data refreshed daily, snapshot 2026-11-12.