Pressure Reveals Character - Scheming: Scheming: leaderboard
Mean 1-5 judge rubric score over 26 multi-turn scenarios testing not engaging in covert goal-directed planning. One behaviour of the Scheming category; a subset of the headline's transcripts.
Metric: Mean judge rubric score over 26 scenarios (1 = failure, 5 = exemplary). Source: storage.googleapis.com. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 4.81 | #176 |
| 2 | Claude Sonnet 4.5 | 4.77 | #250 |
| 3 | Claude Opus 4.5 | 4.73 | #156 |
| 4 | Claude Sonnet 4 | 4.58 | #400 |
| 5 | GPT-5 Mini | 4.54 | #348 |
| 6 | Gemini 3 Pro | 4.27 | #137 |
| 7 | DeepSeek V3.2 | 4.23 | #406 |
| 8 | Llama 4 Maverick | 4.19 | #1099 |
| 9 | Gemini 3 Flash | 4.12 | #164 |
| 10 | Qwen 3 235B A22B | 4.12 | #623 |
| 11 | Mistral Small 3.2 | 4.12 | #1591 |
| 12 | Gemini 2.5 Pro | 4.08 | #266 |
| 13 | Llama 3.1 405B Instruct | 4.08 | #1115 |
| 14 | GPT-4o | 4 | #741 |
| 15 | GPT-OSS-120B | 4 | #700 |
Interactive version: theaggregate.ai/benchmark?slug=pressure-reveals-character-scheming-scheming · How It Works · Data refreshed daily, snapshot 2026-11-13.