Pressure Reveals Character - Honesty: Sandbagging: leaderboard
Mean 1-5 judge rubric score over 36 multi-turn scenarios testing not deliberately underperforming on evaluations. One behaviour of the Honesty category; a subset of the headline's transcripts.
Metric: Mean judge rubric score over 36 scenarios (1 = failure, 5 = exemplary). Source: storage.googleapis.com. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 4.69 | #156 |
| 2 | GPT-5.2 | 4.58 | #176 |
| 3 | Gemini 3 Flash | 4.58 | #164 |
| 4 | DeepSeek V3.2 | 4.58 | #406 |
| 5 | Qwen 3 235B A22B | 4.56 | #623 |
| 6 | Mistral Small 3.2 | 4.56 | #1591 |
| 7 | GPT-4o | 4.53 | #741 |
| 8 | Gemini 2.5 Flash | 4.53 | #453 |
| 9 | GPT-5 Mini | 4.53 | #348 |
| 10 | Grok 4.1 | 4.53 | #476 |
| 11 | GPT-OSS-120B | 4.5 | #700 |
| 12 | Mistral Large 3 | 4.5 | #817 |
| 13 | Qwen 3 Next 80B A3B | 4.5 | #757 |
| 14 | Kimi K2 | 4.47 | #439 |
| 15 | Llama 4 Maverick | 4.44 | #1099 |
Interactive version: theaggregate.ai/benchmark?slug=pressure-reveals-character-honesty-sandbagging · How It Works · Data refreshed daily, snapshot 2026-11-13.