Benchmarks / Reasoning
Blueprint-Bench 2
Spatial reasoning: reconstruct 2D floor plans from ~20 apartment photos, scored on room connectivity and scale (random baseline normalized to 0).
Blueprint-Bench 2
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Gemini 4 | max | 54.4% |
| 2 | Opus 5.5 | max | 51.2% |
| 3 | GPT-6 Astra | max | 49.7% |
| 4 | Fable 5.1 | max | 41.9% |
| 5 | Sonnet 5.5 | max | 39.3% |
| 6 | Fable 5 | max | 38.6% |
| 7 | Gemini 3.8 Flash | max | 38.6% |
| 8 | GPT-6 Sol | max | 36.9% |
| 9 | GPT-5.5 | xhigh | 36.2% |
| 10 | Gemini 3.5 Flash | xhigh | 33.6% |
| 11 | GPT-5.6 Sol | max | 33.6% |
| 12 | Grok 4.6 | max | 33.2% |
| 13 | Haiku 5.5 | max | 33.0% |
| 14 | Grok 4.7 | max | 32.5% |
| 15 | Gemini 3.6 Flash | xhigh | 31.2% |
| 16 | GPT-6 Luna | max | 31.2% |
| 17 | GPT-5.6 Terra | max | 30.8% |
| 18 | Opus 5 | max | 30.4% |
| 19 | Kimi K3 | max | 29.5% |
| 20 | Grok 4.5 | max | 27.3% |
| 21 | GPT-5.4 | xhigh | 27.1% |
| 22 | Gemini 3.1 Pro | high | 26.5% |
| 23 | Sonnet 5 | max | 24.9% |
| 24 | Opus 4.7 | max | 24.5% |
| 25 | GPT-5.6 Luna | max | 22.6% |
| 26 | Opus 4.8 | max | 14.5% |
| 27 | Sonnet 4.6 | max | 6.7% |
| 28 | Kimi K2.6 | max | 3.9% |