LLM Trustworthy - Adversarial: leaderboard
Metric: Trust Score (%). Source: huggingface.co. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3 8B Instruct | 67.28 |
| 2 | Gemini 1.0 Pro | 67.28 |
| 3 | GPT-4 (0314) | 64.04 |
| 4 | Claude 2 | 57.98 |
| 5 | GPT-3.5 Turbo (0301) | 56.69 |
| 6 | vicuna-7B-v1.3 | 52.16 |
| 7 | GPT-4o (2024-05-13) | 51.36 |
| 8 | Llama 2 7B Chat | 51.01 |
| 9 | GPT-4o Mini (2024-07-18) | 50.25 |
| 10 | Mistral-7B-OpenOrca | 47.23 |
| 11 | tulu-2-7B | 44.62 |
| 12 | falcon-7B Instruct | 43.98 |
| 13 | gemma-7B (IT) | 43.43 |
| 14 | gemma-2B (IT) | 43.21 |
| 15 | tulu-2-13B | 43.14 |
Interactive version: theaggregate.ai/benchmark?slug=llm-trustworthy-adversarial · How It Works · Data refreshed daily, snapshot 2026-09-05.