LLM Trustworthy Leaderboard: leaderboard
DecodingTrust safety evaluation across 8 dimensions: toxicity, stereotypes, adversarial robustness, out-of-distribution, privacy, ethics, and fairness.
Metric: Average Trust Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 35 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 2 | 84.52 | #761 |
| 2 | GPT-4o (2024-05-13) | 82.96 | #325 |
| 3 | Llama 3 8B Instruct | 80.61 | #1129 |
| 4 | Gemini 1.0 Pro | 80.61 | #933 |
| 5 | GPT-4o Mini (2024-07-18) | 76.31 | #558 |
| 6 | Llama 2 7B Chat | 74.72 | #1499 |
| 7 | GPT-4 (0314) | 69.24 | #574 |
| 8 | gemma-2B (IT) | 67.18 | #1559 |
| 9 | gemma-7B (IT) | 66.87 | #1496 |
| 10 | tulu-2-13B | 66.51 | #1404 |
| 11 | tulu-2-7B | 63.56 | #1429 |
| 12 | zephyr-7B-beta | 63.24 | #1330 |
| 13 | mpt-7B-chat | 62.29 | #1556 |
| 14 | vicuna-7B-v1.3 | 60.62 | #1507 |
| 15 | falcon-7B Instruct | 59.49 | #1622 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-trustworthy-leaderboard · How It Works · Data refreshed daily, snapshot 2026-10-20.