EPIC-Bench - Visual Matching: leaderboard
Metric: Navigation score (0-100) on visual matching (localizing an object marked in one view in another view of the same scene), on EPIC-Bench (6,661 human-annotated image, text and mask tuples from 25 public datasets across 23 fine-grained embodied perception tasks; the model outputs bounding boxes, counts, path points or feasibility judgements that are scored against the masks), zero-shot, averaged over six runs for local models and two for API models; higher is better. Source: arxiv.org. Saturation forecast: Around December 2027. 89 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 60.08 |
| 2 | Gemini 3.1 Pro (Preview) | 59.06 |
| 3 | GPT-5.5 (Non-reasoning) | 51.65 |
| 4 | Gemini 3 Flash (Preview) | 49.92 |
| 5 | Qwen 3 VL 235B A22B (Thinking) | 49.01 |
| 6 | Gemini 3 Flash (Preview) (Non-reasoning) | 47.25 |
| 7 | Qwen 3.5 27B | 46.77 |
| 8 | GPT-5 Mini | 45.97 |
| 9 | Seed 1.8 | 45.9 |
| 10 | O4 Mini | 45.63 |
| 11 | Qwen 3.5 397B A17B | 45.35 |
| 12 | Qwen 3.5 Plus | 45.21 |
| 13 | Qwen 3.5 122B A10B | 45.19 |
| 14 | GPT-5.1 (Non-reasoning) | 44.69 |
| 15 | GPT-5.4 (Non-reasoning) | 44.35 |
Interactive version: theaggregate.ai/benchmark?slug=epic-bench-visual-matching · How It Works · Data refreshed daily, snapshot 2026-10-07.