360Bench (CMP) - Direction Giving: leaderboard
Metric: Direction-giving accuracy (%) on multi-step route questions toward a visible or signposted target on 360Bench's 1,532 human-written four-option questions on 643 high-resolution (7K) 360-degree images, each scored under four cyclic permutations of the options (6,128 evaluations), greedy decoding, where the image is given as a cubemap projection (CMP) at matching resolution; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 37.2 | #145 |
| 2 | GPT-4o | 35.3 | #333 |
| 3 | Gemini 2.5 Flash | 34.7 | #237 |
| 4 | Qwen 2.5 VL 7B Instruct | 30.6 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=360bench-cmp-direction-giving · How It Works · Data refreshed daily, snapshot 2026-10-11.