Blueprint-Bench
Multimodal · 2025-09-24
Blueprint-Bench evaluates spatial reasoning by asking models and agents to reconstruct 2D apartment floor plans from approximately 20 photographs per apartment. The score is a connectivity-based similarity measure across 50 apartments.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5 | 0.4 |
| Gemini 2.5 Pro | 0.4 |
| GPT-5 Mini | 0.4 |
| Grok 4 | 0.4 |
| Gemini 2.5 Flash | 0.4 |
| Opus 4 | 0.4 |
| Sonnet 4 | 0.3 |
| Gemini 2.5 Flash Image Preview | 0.2 |
| GPT-4o | 0.1 |