SuperGlasses - Multi-Hop: leaderboard
Metric: Accuracy (%) on the 775 questions needing two or more reasoning hops of SuperGlasses' smart-glasses visual question answering set (images captured with Ray-Ban Meta, XiaoMi and RayNeo glasses), direct answers without retrieval, graded against the reference by a Qwen2.5-32B judge; higher is better. Source: arxiv.org. Saturation forecast: Around December 2027. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 38.06 | #145 |
| 2 | GPT-4o | 35.61 | #333 |
| 3 | Claude Sonnet 4 | 30.45 | #194 |
| 4 | Qwen 2.5 VL 72B Instruct | 29.94 | #364 |
| 5 | Qwen 2.5 VL 32B Instruct | 28.13 | #443 |
| 6 | Qwen 2.5 VL 7B Instruct | 24.13 | #643 |
| 7 | InternVL3-8B | 20.26 | #606 |
| 8 | GLM-4.1V-9B (Thinking) | 18.06 | #457 (GLM-4.1V-9B) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=superglasses-multi-hop · How It Works · Data refreshed daily, snapshot 2026-10-11.