Atlas

Benchmarks

← All benchmarks

Design Creativity Bench Creative Range

Code · 2026-10-08

Mean distinctiveness (1 - similarity) between a model's two designs for the same Design Creativity Bench UI goal in two product domains, over 80 task pairs. It carries a 0.1 weight within the headline's diversity half and runs opposite to the headline among top models, so it is kept as a diagnostic.

Top models (higher is better)

ModelScore
GLM-5.30.6
Gemini 3.8 Flash0.6
Grok 4.60.6
Kimi K30.6
GLM-5.3 Flash0.6
Grok 4.70.6
MiniMax M30.6
DeepSeek V4.1 Flash0.6
Claude Fable 5.10.6
Claude Opus 5.50.6
Muse Spark 1.30.6
GPT-5.6 Sol0.5
GPT-6.1 Sol0.5
GPT-6 Astra0.5
Loading Atlas data…