Pressure Reveals Character - Safety: Privacy: leaderboard
Mean 1-5 judge rubric score over 32 multi-turn scenarios testing protecting user privacy and personal information. One behaviour of the Safety category; a subset of the headline's transcripts.
Metric: Mean judge rubric score over 32 scenarios (1 = failure, 5 = exemplary). Source: storage.googleapis.com. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 4.81 | #250 |
| 2 | GPT-5.2 | 4.72 | #176 |
| 3 | Claude Sonnet 4 | 4.53 | #400 |
| 4 | Claude Opus 4.5 | 4.53 | #156 |
| 5 | GPT-5 Mini | 3.91 | #348 |
| 6 | Gemini 2.5 Flash | 3.78 | #453 |
| 7 | DeepSeek V3.2 | 3.75 | #406 |
| 8 | Kimi K2 | 3.56 | #439 |
| 9 | Gemini 3 Pro | 3.47 | #137 |
| 10 | Gemini 2.5 Pro | 3.44 | #266 |
| 11 | GPT-4o | 3.41 | #741 |
| 12 | Qwen 3 235B A22B | 3.34 | #623 |
| 13 | Qwen 3 Next 80B A3B | 3.34 | #757 |
| 14 | GPT-OSS-120B | 3.19 | #700 |
| 15 | Grok 4.1 | 3.13 | #476 |
Interactive version: theaggregate.ai/benchmark?slug=pressure-reveals-character-safety-privacy · How It Works · Data refreshed daily, snapshot 2026-11-13.