Visual Parsing on OmniParsingBench Natural Image module
69.72Overall ScoreGemini-3-Pro
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Gemini-3-Pro2026.03 | 69.72 | 78.33 | 98.64 | 75.69 | 67.15 | 63.15 | 89.87 | 62.7 | 76.29 | |
| Logics-Parsing-Omni2026.03 | 62.46 | 69.58 | 97.88 | 59.81 | 62.49 | 50.53 | 90.36 | 58.4 | 74.38 | |
| Qwen3-VL-235B-A22BParameters=235B, Architecture=A22B2026.03 | 59.34 | 68.66 | 97.85 | 68.61 | 60.63 | 50.62 | 82.7 | 53.4 | 68.05 | |
| Qwen3.5-397B-A17BParameters=397B, Architecture=A17B2026.03 | 54.8 | 63.71 | 97.42 | 80.41 | 52.25 | 47.12 | 76.46 | 48.5 | 62.48 | |
| Qwen3-VL-30B-A3BParameters=30B, Architecture=A3B2026.03 | 53.73 | 67.52 | 97.51 | 57.37 | 59.61 | 46.47 | 75.68 | 46.3 | 60.99 | |
| GPT-5.2Knowledge identification constraints=policy-induced2026.03 | 52.84 | 66.15 | 95.19 | — | 47.2 | 46.07 | 81.72 | 37.5 | 59.61 | |
| Qwen3-Omni-30B-A3BParameters=30B, Architecture=A3B2026.03 | 51.3 | 63.1 | 97.92 | 58.27 | 60.08 | 44.88 | 74.34 | 41.1 | 57.72 |