MedPriv-Bench - Response Utility: leaderboard
Metric: Response utility (0-5): GPT-5-mini grades each answer against question-specific grading points from the gold answer; MedPriv-Bench's 738 test questions: the model answers a clinical question from a retrieved patient record (PMC-Patients summaries) into which identifying or stigmatizing health facts were injected; temperature 0.4; no privacy instruction; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Nano | 4.89 | #415 |
| 2 | DeepSeek V3.2 | 4.72 | #198 |
| 3 | Claude Haiku 4.5 | 4.71 | #271 |
| 4 | Llama3-Med42-70B | 4.54 | #594 |
| 5 | Gemini 3 Flash | 4.14 | #93 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=medpriv-bench-response-utility · How It Works · Data refreshed daily, snapshot 2026-10-11.