Pressure Reveals Character - Safety: leaderboard
Mean 1-5 judge rubric score over the 104 Safety scenarios of Pressure Reveals Character (harmful content, malicious system prompts, privacy and tool-use safety). A subset of the headline's transcripts.
Metric: Mean judge rubric score over 104 scenarios (1 = failure, 5 = exemplary). Source: storage.googleapis.com. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 4.88 | #250 |
| 2 | Claude Opus 4.5 | 4.79 | #156 |
| 3 | GPT-5.2 | 4.77 | #176 |
| 4 | Claude Sonnet 4 | 4.61 | #400 |
| 5 | GPT-5 Mini | 4.29 | #348 |
| 6 | Gemini 2.5 Flash | 4.04 | #453 |
| 7 | Gemini 3 Pro | 4.04 | #137 |
| 8 | DeepSeek V3.2 | 4.02 | #406 |
| 9 | Kimi K2 | 3.92 | #439 |
| 10 | Qwen 3 235B A22B | 3.88 | #623 |
| 11 | Gemini 3 Flash | 3.88 | #164 |
| 12 | Grok 4.1 | 3.85 | #476 |
| 13 | Gemini 2.5 Pro | 3.83 | #266 |
| 14 | GPT-4o | 3.81 | #741 |
| 15 | Qwen 3 Next 80B A3B | 3.81 | #757 |
Interactive version: theaggregate.ai/benchmark?slug=pressure-reveals-character-safety · How It Works · Data refreshed daily, snapshot 2026-11-13.