EditReward-Compass - Visual Consistency: leaderboard
Metric: Preference accuracy (%) on the pairs that differ in visual consistency on EditReward-Compass (2,251 human-verified preference pairs of edited images sampled from the same editing models under the same instruction), general MLLMs prompted with the Edit-Compass rubric; thinking disabled unless the row is a thinking run; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 60.02 |
| 2 | Gemini 3 Flash | 59.81 |
| 3 | Qwen 3.5 27B (Non-reasoning) | 58.5 |
| 4 | Qwen 3.6 27B | 56.56 |
| 5 | Qwen 3.5 27B | 56.37 |
| 6 | Qwen 3.5 35B A3B (Non-reasoning) | 54.79 |
| 7 | Qwen 3.6 35B A3B | 53 |
| 8 | Gemma 4 31B | 51.65 |
| 9 | Qwen 3.5 9B (Non-reasoning) | 50.75 |
| 10 | Qwen 3.5 35B A3B | 50.73 |
| 11 | Qwen 3.6 27B (Non-reasoning) | 49.66 |
| 12 | Qwen 3.5 9B | 48.6 |
| 13 | GPT-4.1 | 48.45 |
| 14 | Qwen 3.6 35B A3B (Non-reasoning) | 45.58 |
| 15 | Gemma 4 26B A4B | 39.6 |
Interactive version: theaggregate.ai/benchmark?slug=editreward-compass-visual-consistency · How It Works · Data refreshed daily, snapshot 2026-10-07.