LLM Trustworthy - Out-of-Distribution: leaderboard
The "Out-of-Distribution" subtask of LLM Trustworthy. AI-Secure's LLM trustworthiness leaderboard (DecodingTrust) — safety and reliability measured across adversarial robustness, ethics, fairness, out-of-distribution robustness, privacy, stereotype and toxicity dimensions, each reported as...
Metric: Trust Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4 (0314) | 87.55 | #574 |
| 2 | GPT-4o (2024-05-13) | 86.59 | #325 |
| 3 | Claude 2 | 85.77 | #761 |
| 4 | GPT-4o Mini (2024-07-18) | 79.07 | #558 |
| 5 | Llama 2 7B Chat | 75.65 | #1499 |
| 6 | Mistral-7B-OpenOrca | 73.4 | #1389 |
| 7 | Llama 3 8B Instruct | 70.85 | #1129 |
| 8 | Gemini 1.0 Pro | 70.85 | #933 |
| 9 | tulu-2-13B | 70.17 | #1404 |
| 10 | tulu-2-7B | 69.3 | #1429 |
| 11 | zephyr-7B-beta | 65.58 | #1330 |
| 12 | mpt-7B-chat | 64.26 | #1556 |
| 13 | gemma-7B (IT) | 61.78 | #1496 |
| 14 | vicuna-7B-v1.3 | 59.1 | #1507 |
| 15 | RedPajama-INCITE-7B-Instruct | 54.21 | #1566 |
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-out-of-distribution · How It Works · Data refreshed daily, snapshot 2026-10-20.