AirGroundBench - Localization Reasoning: leaderboard
Metric: Accuracy (%) on the localization spatial-reasoning VQA task; higher is better. Source: arxiv.org. Saturation forecast: Around 2031. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Seed 2.0 Pro | 66.23 |
| 2 | Qwen 3.5 397B A17B | 57.12 |
| 3 | Qwen 3.5 Plus | 56.76 |
| 4 | GLM-4.6V | 56.76 |
| 5 | Claude Sonnet 4.6 | 56.36 |
| 6 | Qwen 3.5 27B | 56.32 |
| 7 | Kimi K2.5 | 55.41 |
| 8 | ERNIE 5.0 | 44.74 |
| 9 | GPT-5.2 | 43.24 |
| 10 | Gemini 3.1 Pro (Preview) | 41.11 |
| 11 | Qwen 3 VL 235B A22B | 40.54 |
Interactive version: theaggregate.ai/benchmark?slug=airgroundbench-localization-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-29.