PSEBench: leaderboard

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

#ModelScore
1Gemini 3.1 Pro (Preview)95.3
2GPT-5.594.6
3Claude Opus 4.792.3
4GPT-590.2
5Gemini 2.5 Flash85.8
6Claude Sonnet 4.684.5
7DeepSeek R183.7
8GPT-OSS-120B79
9Qwen 3 235B A22B 2507 Instruct76.7
10GPT-5 Nano71.9
11Mistral Small 3.260.5
12Llama 3.1 8B Instruct52.4

Interactive version: theaggregate.ai/benchmark?slug=psebench · How It Works · Data refreshed daily, snapshot 2026-09-29.