EPAG - Weighted HPI Guideline Score: leaderboard
Metric: Guideline-Matched History Units per Dialogue (high=2, medium=1). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 | 8.12 |
| 2 | Claude 3.7 Sonnet | 8.12 |
| 3 | Claude 3.5 Sonnet | 8.05 |
| 4 | Claude 3.5 Haiku | 7.84 |
| 5 | GPT-4o Mini | 7.75 |
| 6 | GPT-4.1 Mini | 7.64 |
| 7 | GPT-4o | 7.59 |
| 8 | Phi-3.5-mini-instruct | 6.88 |
Interactive version: theaggregate.ai/benchmark?slug=epag-weighted-hpi-guideline-score · How It Works · Data refreshed daily, snapshot 2026-09-25.