In-context image editing on MICE-Bench Turn 3
62.06IFGPT-Image-1 [high]
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-Image-1 [high]2026.06 | 62.06 | 60.55 | 52.76 | 58.46 | |
| Edit-R22026.06 | 58.63 | 78.51 | 58.96 | 65.37 | |
| Nano-Banana-22026.06 | 57.98 | 76.37 | 52.43 | 62.26 | |
| BAGELvariant=w/ IC-CoT2026.06 | 50.82 | 78.21 | 46.79 | 58.61 | |
| BAGEL2026.06 | 50.18 | 79.9 | 43.59 | 57.89 | |
| BAGELvariant=w/ think2026.06 | 47.8 | 77.88 | 48.14 | 57.94 | |
| VINCIE2026.06 | 43.78 | 58.02 | 26.9 | 42.9 | |
| OmniGen2026.06 | 40.97 | 47.41 | 36.76 | 41.71 | |
| Flux.1-Kontext [dev]Originally support in-context editing=false2026.06 | 33.9 | 68.53 | 31.87 | 44.77 | |
| Qwen-Image-Edit [2511]Originally support in-context editing=false2026.06 | 25.04 | 72.46 | 20.76 | 39.42 |