ArchSIBench: leaderboard
Architectural spatial intelligence, the ability to recognize and infer architectural space, is fundamental to tasks such as robot navigation, embodied interaction, and 3D scene understanding and generation. Although extensive research has evaluated the basic spatial skills of Vision-Language Models (VLMs) such as relative orientation, distance comparison, and object counting, these tasks cover onl
Metric: Avg. (self-reported). Source: benchmarklist.com. Saturation forecast: Around June 2027. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Best Expert | 92.3 | |
| 2 | Median Expert | 89.2 | |
| 3 | Median Human | 85.1 | |
| 4 | Gemini 3 Pro | 77.9 | #80 |
| 5 | Gemini 3 Flash | 77.6 | #99 |
| 6 | Gemini 3.1 Pro (Preview) | 77.2 | #58 |
| 7 | Qwen 3.5 Plus | 63.2 | #125 |
| 8 | GPT-5 Mini | 62.8 | #186 |
| 9 | Gemma 4 31B (IT) | 62.5 | #176 |
| 10 | Qwen 3.5 397B A17B | 56.7 | #146 |
| 11 | Qwen 3.5 27B | 55.9 | #187 |
| 12 | GPT-5.4 | 53.8 | #79 |
| 13 | Qwen 3.5 122B A10B | 53.6 | #173 |
| 14 | GPT-5.2 | 53.5 | #110 |
| 15 | Claude Opus 4.6 | 53.3 | #63 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=archsibench · How It Works · Data refreshed daily, snapshot 2026-10-20.