MultihopSpatial: leaderboard
Metric: Multiple-choice accuracy (%) over all 4,500 MultihopSpatial questions (1-3 hops, ego- and exo-centric), one prompt template, answer and target bounding box required; higher is better. Source: arxiv.org. Saturation forecast: Around August 2028. 37 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 64.7 | #77 |
| 2 | GPT-5.2 (xHigh) | 57.9 | #105 (GPT-5.2) |
| 3 | Gemini 3 Flash | 57.2 | #93 |
| 4 | Claude Opus 4.5 (Thinking) | 47 | #79 (Claude Opus 4.5) |
| 5 | Qwen 3 VL 32B (Thinking) | 46.8 | #287 (Qwen 3 VL 32B) |
| 6 | Claude Opus 4.5 | 45.1 | #79 |
| 7 | Qwen 3 VL 235B A22B (Thinking) | 45.1 | #228 (Qwen 3 VL 235B A22B) |
| 8 | GLM-4.6V (Non-reasoning) | 43.2 | #309 (GLM-4.6V) |
| 9 | Qwen 3 VL 4B (Thinking) | 42.6 | #471 (Qwen 3 VL 4B) |
| 10 | GLM-4.6V (Thinking) | 42 | #309 (GLM-4.6V) |
| 11 | Molmo2-8B | 41.8 | #522 |
| 12 | Qwen 3 VL 8B (Thinking) | 41.7 | |
| 13 | Qwen 3 VL 235B A22B Instruct | 41.3 | #264 |
| 14 | Qwen 3 VL 32B Instruct | 40.9 | #276 |
| 15 | Qwen 3 VL 8B Instruct | 38 | #401 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=multihopspatial · How It Works · Data refreshed daily, snapshot 2026-10-11.