LLM Trustworthy - Toxicity — leaderboard
Metric: Trust Score (%). Source: huggingface.co. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 2 | 92.11 |
| 2 | GPT-4o (2024-05-13) | 86.46 |
| 3 | Llama 2 7B Chat | 80 |
| 4 | Llama 3 8B Instruct | 77.53 |
| 5 | Gemini 1.0 Pro | 77.53 |
| 6 | gemma-2B (IT) | 77.07 |
| 7 | gemma-7B (IT) | 75.52 |
| 8 | GPT-4o Mini (2024-07-18) | 59.02 |
| 9 | GPT-3.5 Turbo (0301) | 47 |
| 10 | tulu-2-13B | 44.8 |
| 11 | GPT-4 (0314) | 41 |
| 12 | mpt-7B-chat | 40 |
| 13 | falcon-7B Instruct | 39 |
| 14 | zephyr-7B-beta | 31.97 |
| 15 | Mistral-7B-OpenOrca | 30.12 |
Interactive version: theaggregate.ai/benchmark?slug=llm-trustworthy-toxicity · How the rankings work · Data refreshed daily, snapshot 2026-07-22.