Model Vulnerability Evaluation on TRAP Image 1.0
76.5Task Performance ScoreQwen3-VL-8B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-VL-8BModel Type=Open-source2026.06 | 76.5 | 15 | |
| Qwen3-VL-32BModel Type=Open-source2026.06 | 75.5 | 13.5 | |
| GPT-5-miniModel Type=Proprietary2026.06 | 71.5 | 16 | |
| Devstral-Small-2512Model Type=Open-source2026.06 | 71.5 | 14.5 | |
| GPT-5.4-miniModel Type=Proprietary2026.06 | 71 | 17.5 | |
| Gemini-2.5-FlashModel Type=Proprietary2026.06 | 67.5 | 15 | |
| GPT-5Model Type=Proprietary2026.06 | 64.5 | 16 | |
| GPT-4o-miniModel Type=Proprietary2026.06 | 64 | 18 | |
| InternVL3.5-14BModel Type=Open-source2026.06 | 64 | 20 | |
| InternVL3.5-38BModel Type=Open-source2026.06 | 63 | 16.5 | |
| Gemini-2.5-ProModel Type=Proprietary2026.06 | 62 | 28.5 | |
| InternVL3.5-8BModel Type=Open-source2026.06 | 61 | 16.5 | |
| Claude Sonnet 4.5Model Type=Proprietary2026.06 | 60 | 12.5 | |
| Gemini-2.5-Flash-LiteModel Type=Proprietary2026.06 | 57 | 26.5 | |
| Qwen3-VL-4BModel Type=Open-source2026.06 | 56 | 41.5 | |
| Claude Haiku 4.5Model Type=Proprietary2026.06 | 55.5 | 12.5 | |
| InternVL3.5-2BModel Type=Open-source2026.06 | 52 | 19 | |
| Phi-4-multimodalModel Type=Open-source2026.06 | 30.5 | 44 | |
| GLM-4.6V-FlashModel Type=Open-source2026.06 | 28 | 21 | |
| InternVL3.5-4BModel Type=Open-source2026.06 | 27 | 17.5 | |
| Llama3.2-11B-VisionModel Type=Open-source2026.06 | 20.5 | 18 | |
| Qwen3-VL-2BModel Type=Open-source2026.06 | 11 | 29.5 |