Situational Awareness Dataset (SAD) (Situating Prompt): leaderboard
Metric: Score (%, higher is better). Source: situational-awareness-dataset.org. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O1 Preview (2024-09-12) | 61.93 |
| 2 | Claude 3.5 Sonnet | 57.13 |
| 3 | O1 Mini (2024-09-12) | 55.37 |
| 4 | Claude 3 Opus | 53.37 |
| 5 | GPT-4 Preview (0125) | 51.52 |
| 6 | GPT-4o | 50.72 |
| 7 | Claude 3 Sonnet | 50.01 |
| 8 | Llama 3 70B Chat | 49.34 |
| 9 | GPT-4 (0613) | 46.73 |
| 10 | Claude 2.1 | 46.19 |
| 11 | Claude 3 Haiku | 45.88 |
| 12 | Claude Instant 1.2 | 43.9 |
| 13 | Llama 2 70B Chat | 40.23 |
| 14 | GPT-4 Base | 39.29 |
| 15 | GPT-3.5 Turbo (0613) | 38.42 |
Interactive version: theaggregate.ai/benchmark?slug=situational-awareness-dataset-sad-situating-prompt · How It Works · Data refreshed daily, snapshot 2026-09-19.