EPAG - HPI Guideline Score: leaderboard
Metric: Guideline-Matched History Units per Dialogue (unweighted). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 | 4.82 |
| 2 | Claude 3.5 Sonnet | 4.62 |
| 3 | Claude 3.7 Sonnet | 4.59 |
| 4 | Claude 3.5 Haiku | 4.58 |
| 5 | GPT-4.1 Mini | 4.46 |
| 6 | GPT-4o Mini | 4.46 |
| 7 | GPT-4o | 4.39 |
| 8 | Phi-3.5-mini-instruct | 3.91 |
Interactive version: theaggregate.ai/benchmark?slug=epag-hpi-guideline-score · How It Works · Data refreshed daily, snapshot 2026-09-25.