Long-document Visual Question Answering on MMLongBench Overall
90.77Average ScoreGPT-5.5
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-5.5Model Category=Closed-source LVLMs2026.05 | 90.77 | |
| Gemini-3.1-ProModel Category=Closed-source LVLMs2026.05 | 83.66 | |
| Gemini-3.1-FlashModel Category=Closed-source LVLMs2026.05 | 80.13 | |
| Gemini-2.5-ProModel Category=Closed-source LVLMs2026.05 | 74.37 | |
| GPT-5.4Model Category=Closed-source LVLMs2026.05 | 69.41 | |
| Gemma4-31BModel Category=Large open-source LVLMs (≥ 15B), Parameter Scale=31B2026.05 | 67.1 | |
| Qwen2.5-VL-72BModel Category=Large open-source LVLMs (≥ 15B), Parameter Scale=72B2026.05 | 60.83 | |
| Qwen2.5-VL-32BModel Category=Large open-source LVLMs (≥ 15B), Parameter Scale=32B2026.05 | 58.31 | |
| MMProLongModel Category=Small open-source LVLMs (< 15B)2026.05 | 57.7 | |
| Gemma3-27BModel Category=Large open-source LVLMs (≥ 15B), Parameter Scale=27B2026.05 | 52.63 | |
| Qwen2.5-VL-7BModel Category=Small open-source LVLMs (< 15B), Parameter Scale=7B2026.05 | 50.59 | |
| InternVL3-38BModel Category=Large open-source LVLMs (≥ 15B), Parameter Scale=38B2026.05 | 48.88 | |
| Gemma3-12BModel Category=Small open-source LVLMs (< 15B), Parameter Scale=12B2026.05 | 47.76 | |
| Gemma4-26B-A4BModel Category=Large open-source LVLMs (≥ 15B), Parameter Scale=26B-A4B2026.05 | 47.74 | |
| InternVL3-14BModel Category=Small open-source LVLMs (< 15B), Parameter Scale=14B2026.05 | 47.47 | |
| InternVL3-8BModel Category=Small open-source LVLMs (< 15B), Parameter Scale=8B2026.05 | 47.15 | |
| Gemma3-4BModel Category=Small open-source LVLMs (< 15B), Parameter Scale=4B2026.05 | 38.03 | |
| Gemma4-E4BModel Category=Small open-source LVLMs (< 15B), Parameter Scale=E4B2026.05 | 37.08 | |
| Gemma4-E2BModel Category=Small open-source LVLMs (< 15B), Parameter Scale=E2B2026.05 | 34.24 | |
| InternVL3.5-38BModel Category=Large open-source LVLMs (≥ 15B), Parameter Scale=38B2026.05 | 33.81 | |
| InternVL3.5-14BModel Category=Small open-source LVLMs (< 15B), Parameter Scale=14B2026.05 | 30.23 | |
| InternVL3.5-8BModel Category=Small open-source LVLMs (< 15B), Parameter Scale=8B2026.05 | 27.56 |