SEAL - MASK: leaderboard
Consistency-based benchmark measuring LLM honesty under pressure, testing whether models contradict their own beliefs when incentivized to deceive across 500 scenarios.
Metric: Score. Source: scale.com. Saturation forecast: Not forecast. 67 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 (Non-reasoning) | 96.28 | #63 (Claude Opus 4.6) |
| 2 | Claude Sonnet 4.5 (Thinking) | 96.13 | #142 (Claude Sonnet 4.5) |
| 3 | Claude Sonnet 4 (Thinking) | 95.33 | #199 (Claude Sonnet 4) |
| 4 | Claude Opus 4.1 (20250805) (Thinking) | 94.2 | #132 (Claude Opus 4.1 (20250805)) |
| 5 | Claude Opus 4.5 (20251101) (Thinking) | 92.53 | #74 (Claude Opus 4.5 (20251101)) |
| 6 | GPT-OSS-120B | 92 | #341 |
| 7 | Claude Sonnet 4 | 89.27 | #199 |
| 8 | Claude Opus 4 (Thinking) | 87.87 | #163 (Claude Opus 4) |
| 9 | Claude Opus 4.1 (20250805) | 87.4 | #132 |
| 10 | Claude Opus 4.5 (20251101) | 87.13 | #74 |
| 11 | GPT-5.2 | 86.67 | #110 |
| 12 | GPT-OSS-20B | 86.46 | #518 |
| 13 | Claude Sonnet 4.5 | 86.4 | #142 |
| 14 | GPT-5.1 (Thinking) | 86.33 | #135 (GPT-5.1) |
| 15 | Claude Opus 4.6 (Adaptive Reasoning, Max Effort) | 85.4 | #63 (Claude Opus 4.6) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=seal-mask · How It Works · Data refreshed daily, snapshot 2026-10-20.