LLM Stats (RefSpatialBench): leaderboard
Referential spatial reasoning benchmark measuring ability to understand spatial relationships and locate objects using reference descriptions in visual contexts.
Metric: Score (%). Source: llm-stats.com. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen3.6-27B | 70 | |
| 2 | Qwen3 VL 235B A22B Thinking | 69.9 | |
| 3 | Qwen3.5-122B-A10B | 69.3 | |
| 4 | Qwen3.5-27B | 67.7 | |
| 5 | Qwen3.6-35B-A3B | 64.3 | |
| 6 | Qwen3.5-35B-A3B | 63.5 | |
| 7 | Seed 1.8 | 56.3 | #139 |
| 8 | Seed 2.0 Mini | 55.6 | #276 |
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-refspatialbench · How It Works · Data refreshed daily, snapshot 2026-10-20.