Deception Resistance (Lechmazur) — leaderboard
Measures LLM resilience to misleading arguments: models must maintain accurate reasoning when exposed to persuasive disinformation from other models.
Metric: Vulnerability Score (lower is better). Source: github.com. Status: saturation imminent. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3 Opus | 0.28 |
| 2 | Claude 3.5 Sonnet | 0.28 |
| 3 | O1 Preview | 0.32 |
| 4 | Mistral Large 2 (Jul) | 0.35 |
| 5 | O1 Mini | 0.5 |
| 6 | Llama 3.1 405B | 0.54 |
| 7 | Qwen 2.5 72B | 0.61 |
| 8 | GPT-4o | 0.61 |
| 9 | Gemini 1.5 Pro (Sept) | 0.62 |
| 10 | Gemini 1.5 Flash | 0.64 |
| 11 | Claude 3 Haiku | 0.78 |
| 12 | Grok 2 | 0.81 |
| 13 | Llama 3.1 70B | 0.82 |
| 14 | GPT-4o Mini | 1.07 |
| 15 | GPT-4 Turbo | 1.18 |
Interactive version: theaggregate.ai/benchmark?slug=deception-resistance-lechmazur · How the rankings work · Data refreshed daily, snapshot 2026-07-22.