Pressure Reveals Character - Honesty: Sycophancy: leaderboard
Mean 1-5 judge rubric score over 41 multi-turn scenarios testing not agreeing with the user merely to please them. One behaviour of the Honesty category; a subset of the headline's transcripts.
Metric: Mean judge rubric score over 41 scenarios (1 = failure, 5 = exemplary). Source: storage.googleapis.com. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 4.68 | #250 |
| 2 | Claude Opus 4.5 | 4.46 | #156 |
| 3 | GPT-5.2 | 4.22 | #176 |
| 4 | GPT-OSS-120B | 3.51 | #700 |
| 5 | GPT-5 Mini | 3.41 | #348 |
| 6 | Kimi K2 | 3.24 | #439 |
| 7 | Gemini 3 Pro | 3.17 | #137 |
| 8 | DeepSeek V3.2 | 3.15 | #406 |
| 9 | Grok 4.1 | 3.15 | #476 |
| 10 | Claude Sonnet 4 | 3.12 | #400 |
| 11 | Gemini 2.5 Flash | 3.05 | #453 |
| 12 | Llama 4 Maverick | 3.02 | #1099 |
| 13 | Gemini 3 Flash | 3 | #164 |
| 14 | Qwen 3 235B A22B | 2.88 | #623 |
| 15 | Llama 3.3 70B | 2.88 | #1498 |
Interactive version: theaggregate.ai/benchmark?slug=pressure-reveals-character-honesty-sycophancy · How It Works · Data refreshed daily, snapshot 2026-11-13.