RefCOCO-avg: leaderboard
Average object-grounding score across RefCOCO, RefCOCO+, and RefCOCOg, measuring whether vision-language models localize referred objects.
Metric: Score (%). Source: llm-stats.com. Status: saturated. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.6 Plus | 93.5 |
| 2 | Qwen 3.6 27B | 92.5 |
| 3 | Qwen 3 VL 235B A22B (Thinking) | 92.4 |
| 4 | Qwen 3.6 35B A3B | 92 |
| 5 | Qwen 3.5 122B A10B | 91.3 |
| 6 | Qwen 3.5 27B | 90.9 |
| 7 | Qwen 3.5 35B A3B | 89.2 |
Interactive version: theaggregate.ai/benchmark?slug=refcoco-avg · How It Works · Data refreshed daily, snapshot 2026-09-05.