DiCoBench - Entity Difference: leaderboard
Metric: Accuracy (%) on the entity-difference task (a small object added, removed or replaced) of DiCoBench (765 multi-image samples at near-2K resolution): 5-option multiple choice (four candidate image pairs or instructions plus a no-visible-difference option, chance 20%), answer letter matched exactly, VLMEvalKit settings at temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 89.5 |
| 2 | GPT-4.1 Mini | 55.8 |
| 3 | Gemini 3 Flash | 55.6 |
| 4 | GPT-5 | 54.7 |
| 5 | Qwen 2.5 VL 7B Instruct | 53.7 |
| 6 | O4 Mini | 53.7 |
| 7 | GPT-4o | 46.3 |
| 8 | Qwen 3.5 35B A3B | 42.1 |
| 9 | Qwen 2.5 VL 32B Instruct | 35.8 |
| 10 | Gemma 3 12B (IT) | 34.7 |
| 11 | Gemma 3 27B (IT) | 30.5 |
| 12 | GPT-4.1 | 30.5 |
Interactive version: theaggregate.ai/benchmark?slug=dicobench-entity-difference · How It Works · Data refreshed daily, snapshot 2026-09-29.