Design Creativity Bench Originality
Code · 2026-10-08
Mean embedding distinctiveness (1 - similarity) between a model's design and other labs' panel designs for the same Design Creativity Bench prompt, over 160 prompts. It is measured against the paper's 15-model panel, with same-lab pairs left out. It is a diagnostic because distinctiveness depends on which models are in the panel.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 0.6 |
| Grok 4.7 | 0.6 |
| GPT-6.1 Sol | 0.6 |
| Gemini 3.8 Flash | 0.6 |
| Grok 4.6 | 0.6 |
| GPT-5.6 Sol | 0.6 |
| MiniMax M3 | 0.6 |
| GLM-5.3 | 0.6 |
| DeepSeek V4.1 Flash | 0.6 |
| Muse Spark 1.3 | 0.6 |
| Claude Fable 5.1 | 0.6 |
| Claude Opus 5.5 | 0.6 |
| GLM-5.3 Flash | 0.6 |
| Kimi K3 | 0.5 |