PSEBench: leaderboard
PSEBench (2026) tests patient safety event triage: from an event narrative and the jurisdiction policy, decide whether the event is Reportable, Non-Reportable or Uncertain. Scored by three-way verdict accuracy over 5,074 complete, missing-information and gray-zone cases, in percent; higher is better.
Metric: Verdict accuracy (%, M1): three-way triage accuracy (Reportable, Non-Reportable, Uncertain) over all cases, parse failures counted wrong, on the 5,074-case MN29 benchmark (3,455 complete, 1,362 missing-information and 257 uncertain cases built from clause cards of the Minnesota 29 reportable adverse health events), agentic environment where the model may ASK a GPT-5.2 information provider within 10 turns and then answers with a verdict, clause, legal evidence and rationale; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 95.3 | #94 |
| 2 | GPT-5.5 | 94.6 | #49 |
| 3 | Claude Opus 4.7 | 92.3 | #79 |
| 4 | GPT-5 | 90.2 | #167 |
| 5 | Gemini 2.5 Flash | 85.8 | #453 |
| 6 | Claude Sonnet 4.6 | 84.5 | #161 |
| 7 | DeepSeek R1 | 83.7 | #473 |
| 8 | GPT-OSS-120B | 79 | #700 |
| 9 | Qwen 3 235B A22B 2507 Instruct | 76.7 | #619 |
| 10 | GPT-5 Nano | 71.9 | #889 |
| 11 | Mistral Small 3.2 | 60.5 | #1591 |
| 12 | Llama 3.1 8B Instruct | 52.4 | #4246 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=psebench · How It Works · Data refreshed daily, snapshot 2026-11-13.