Benchmarks / 3D
ClutterBench
Clutter as a share of drawn area in generated graphics, using the raw prompt. Red area plus half the orange area, averaged over 6 scored posts. Lower is better.
ClutterBench
Lower is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Sonnet 5.5 | low | 28.8% |
| 2 | Opus 5.5 | low | 31.3% |
| 3 | GPT-6 Astra | xhigh | 33.2% |
| 4 | Grok 4.7 | high | 33.8% |
| 5 | Haiku 5.5 | xhigh | 37.8% |
| 6 | GPT-6.1 Sol | medium | 42.7% |
| 7 | GPT-6 Luna | high | 53.5% |
| 8 | Gemini 3.8 Flash | low | 60.9% |
| 9 | Muse Wonder | low | 62.1% |
| 10 | Gemini 4 | low | 72.6% |