COMPL-AI - AI Identity Disclosure: leaderboard
The model is asked yes/no questions about whether it is human; the score is the share answered NO. It is COMPL-AI's EU AI Act transparency test for disclosing being an AI.
Metric: Disclosure of being an AI (%, higher is better). Source: huggingface.co. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral 7B Instruct (v0.3) | 100 | #6653 |
| 2 | Claude 3 Opus | 100 | #1455 |
| 3 | Llama 2 13B Chat | 100 | #8186 |
| 4 | Gemini 1.5 Flash (001) | 100 | #1784 |
| 5 | Gemma 2 9B | 98.63 | #2802 |
| 6 | Mistral 7B Instruct (v0.2) | 98.63 | #7007 |
| 7 | GPT-4 Turbo (Preview) | 97.26 | #1083 |
| 8 | Mistral-7B-v0.3 | 94.52 | #7191 |
| 9 | Llama 2 7B Chat | 93.15 | #8805 |
| 10 | Mixtral 8x7B Instruct | 89.04 | #4444 |
| 11 | Llama 2 70B Chat | 89.04 | #6800 |
| 12 | Qwen 1.5 72B Chat | 72.6 | #2126 |
| 13 | Bielik-11B-v2.3-Instruct | 67.12 | #1791 |
| 14 | GPT-3.5 Turbo (0125) | 58.9 | #3500 |
| 15 | Yi Chat 34B | 35.62 | #3544 |
Interactive version: theaggregate.ai/benchmark?slug=compl-ai-ai-identity-disclosure · How It Works · Data refreshed daily, snapshot 2026-11-13.