Pressure Reveals Character - Safety: Harmful Content: leaderboard
Mean 1-5 judge rubric score over 33 multi-turn scenarios testing refusing to generate content that could cause harm. One behaviour of the Safety category; a subset of the headline's transcripts.
Metric: Mean judge rubric score over 33 scenarios (1 = failure, 5 = exemplary). Source: storage.googleapis.com. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 5 | #156 |
| 2 | Claude Sonnet 4.5 | 4.91 | #250 |
| 3 | Gemini 3 Flash | 4.88 | #164 |
| 4 | Claude Sonnet 4 | 4.88 | #400 |
| 5 | Qwen 3 235B A22B | 4.88 | #623 |
| 6 | Gemini 2.5 Flash | 4.85 | #453 |
| 7 | GPT-5.2 | 4.85 | #176 |
| 8 | Grok 4.1 | 4.85 | #476 |
| 9 | GPT-5 Mini | 4.82 | #348 |
| 10 | GPT-4o | 4.79 | #741 |
| 11 | Gemini 2.5 Pro | 4.79 | #266 |
| 12 | DeepSeek V3.2 | 4.79 | #406 |
| 13 | Gemini 3 Pro | 4.79 | #137 |
| 14 | Qwen 3 14B | 4.76 | #1233 |
| 15 | Qwen 3 Next 80B A3B | 4.67 | #757 |
Interactive version: theaggregate.ai/benchmark?slug=pressure-reveals-character-safety-harmful-content · How It Works · Data refreshed daily, snapshot 2026-11-13.