ParsiEval (378 Questions): leaderboard
A ParsiEval run scored over a 378-question edition rather than the 364-question set; published separately because it is not the same exam. Accuracy, higher is better. Source: github.com/mshojaei77/ParsiEval.
Metric: Accuracy (%, 378 questions). Source: github.com. 1 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | gpt-5-2025-08-07 | 82.01 |
Interactive version: theaggregate.ai/benchmark?slug=parsieval-378-questions · How It Works · Data refreshed daily, snapshot 2026-11-05.