MultihopSpatial - 3-Hop Ego-Centric: leaderboard
Metric: Multiple-choice accuracy (%) on the 3-hop ego-centric questions of MultihopSpatial; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 37 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 39.7 | #77 |
| 2 | GPT-5.2 (xHigh) | 36.1 | #105 (GPT-5.2) |
| 3 | Claude Opus 4.5 | 25.7 | #79 |
| 4 | Claude Opus 4.5 (Thinking) | 25.5 | #79 (Claude Opus 4.5) |
| 5 | Gemma 3 4B (IT) | 22.1 | #971 |
| 6 | Qwen 3 VL 4B (Thinking) | 20.4 | #471 (Qwen 3 VL 4B) |
| 7 | Qwen 3 VL 32B (Thinking) | 19.2 | #287 (Qwen 3 VL 32B) |
| 8 | Qwen 3 VL 8B (Thinking) | 18.5 | |
| 9 | Gemma 3 27B (IT) | 18.1 | #509 |
| 10 | Qwen 3 VL 235B A22B (Thinking) | 17.6 | #228 (Qwen 3 VL 235B A22B) |
| 11 | Gemma 3 12B (IT) | 16.9 | #655 |
| 12 | Molmo2-8B | 15.9 | #522 |
| 13 | GLM-4.6V (Non-reasoning) | 15.9 | #309 (GLM-4.6V) |
| 14 | Qwen 3 VL 4B Instruct | 15.5 | #506 |
| 15 | Qwen 3 VL 235B A22B Instruct | 14.8 | #264 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=multihopspatial-3-hop-ego-centric · How It Works · Data refreshed daily, snapshot 2026-10-11.