SAD - Anti-Imitation (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) | 29 |
| 2 | Claude 3.5 Sonnet | 26.2 |
| 3 | Llama 2 13B | 20.59 |
| 4 | GPT-4 Base | 16.82 |
| 5 | davinci-002 | 15.39 |
| 6 | Llama 2 70B | 13.36 |
| 7 | Claude 3 Opus | 12.79 |
| 8 | GPT-4 Preview (0125) | 11.16 |
| 9 | GPT-4o | 10.88 |
| 10 | Claude 2.1 | 9.75 |
| 11 | Llama 2 7B | 9.07 |
| 12 | O1 Mini (2024-09-12) | 8 |
| 13 | GPT-4 (0613) | 7.55 |
| 14 | GPT-3.5 Turbo (0613) | 6.73 |
| 15 | Claude 3 Sonnet | 5.59 |
Interactive version: theaggregate.ai/benchmark?slug=sad-anti-imitation-situating-prompt · How It Works · Data refreshed daily, snapshot 2026-09-19.