Global Perception on SPD-Faith Bench Multi-Difference 1.0 (test)
68.3DQR ScoreDeepSeek-VL2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DeepSeek-VL2Access Type=Open-Source2026.02 | 68.3 | 62.8 | |
| GLM-4.5VAccess Type=Open-Source2026.02 | 67 | 92.5 | |
| Qwen2.5-VL-7BAccess Type=Open-Source2026.02 | 65 | 42.8 | |
| Qwen3-VL-235B-A22BAccess Type=Open-Source2026.02 | 62.8 | 90.1 | |
| InternVL2.5-38BAccess Type=Open-Source2026.02 | 62.8 | 65.7 | |
| GPT-4oAccess Type=Close-Source2026.02 | 62.2 | 82 | |
| Qwen2.5-VL-72BAccess Type=Open-Source2026.02 | 61.6 | 44.4 | |
| Qwen3-VL-32BAccess Type=Open-Source2026.02 | 60.8 | 67.2 | |
| Gemini-2.5-ProAccess Type=Close-Source2026.02 | 57.3 | 79.7 | |
| Claude-4.5-HaikuAccess Type=Close-Source2026.02 | 54.2 | 64.5 | |
| InternVL2.5-8BAccess Type=Open-Source2026.02 | 54 | 49.5 | |
| MiniCPM-V-2.6Access Type=Open-Source2026.02 | 49.6 | 31.2 |