MentalBench (DSM Diagnosis) - Ambiguous Differential Diagnosis: leaderboard
Metric: Accuracy (%) on the 6,525 Type 3 cases (cases whose discriminating evidence is withheld, so both diagnoses of a differential pair are correct; choose one or more); cases synthesized from a psychiatrist-built DSM-5 knowledge graph by three LLMs and validated by a psychiatrist and a clinical psychologist; four options per case; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 8B | 55.4 | #667 |
| 2 | Qwen 3 235B A22B | 54.19 | #304 |
| 3 | Qwen 2.5 72B Instruct | 52.75 | #436 |
| 4 | Gemma 3 27B (IT) | 52.51 | #509 |
| 5 | Qwen 3 14B | 52.26 | #524 |
| 6 | Qwen 2.5 7B Instruct | 50.7 | #846 |
| 7 | Qwen 2.5 32B Instruct | 49.2 | #491 |
| 8 | Gemma 3 12B (IT) | 48.28 | #655 |
| 9 | Qwen 3 32B | 48.03 | #424 |
| 10 | Qwen 2.5 14B Instruct | 44.78 | #634 |
| 11 | GPT-4o | 40.25 | #333 |
| 12 | Llama 3.1 8B Instruct | 40.12 | #1018 |
| 13 | Llama 3.1 70B Instruct | 39.75 | #548 |
| 14 | Gemini 2.5 Pro | 35.16 | #145 |
| 15 | Gemma 3 4B (IT) | 34.04 | #971 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=mentalbench-dsm-diagnosis-ambiguous-differential-diagnosis · How It Works · Data refreshed daily, snapshot 2026-10-11.