Unified image generation and editing on ICE-Bench (test)
5.801Aesthetics ScoreGPT-4o
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT-4oModel Type=close-source produce-level2026.01 | 5.801 | 82.3 | 27.8 | 69.3 | 88.9 | |
| Gemini-2.5-flashModel Type=close-source produce-level2026.01 | 5.571 | 87.9 | 28.1 | 72.4 | 84.7 | |
| TAG-MoEModel Type=open-source2026.01 | 5.399 | 85.7 | 28.2 | 73.2 | 85.2 | |
| Qwen-EditModel Type=open-source2026.01 | 5.358 | 84 | 27.9 | 67.1 | 77.4 | |
| OmniGen2Model Type=open-source2026.01 | 5.238 | 85.5 | 27.9 | 72.8 | 78.7 | |
| ACE++Model Type=open-source2026.01 | 5.219 | 85.1 | 26.3 | 71.3 | 63.7 | |
| DreamOmni2Model Type=open-source2026.01 | 5.188 | 86.6 | 26.8 | 73.9 | 66.4 | |
| KontextModel Type=open-source2026.01 | 5.165 | 86.3 | 27.4 | 72.8 | 62.9 | |
| BAGELModel Type=open-source2026.01 | 4.757 | 86.3 | 27.6 | 68.7 | 69.9 |