Name: LLM Divergent Thinking Score
Published: 2025-03-20

LLM Divergent Thinking Score

Chat & Writing · 2025-03-20

Lech Mazur's Divergent Thinking Creativity Benchmark asks models to generate many unique words under letter and association constraints. The headline score rewards semantic distance and valid diversity.

Top models (higher is better)

Model	Score
o1 Preview	4.8
Gemini 2.0 Flash Experimental	4.7
Opus 3	4.5
Grok 2	4.5
Llama 3.3 70B Instruct	4.4
Gemini 2.0 Flash Thinking Experimental 01-21	4.4
Claude 3.5 Sonnet (Oct 2024)	4.4
Gemma 2 27B	4.4
o1-mini	4.2
Haiku 3.5	4.2
Mistral Large 2 (Instruct 2407)	4.1
GPT-4o Mini	4.1
Gemini 1.5 Flash	4.1
Gemini 1.5 Pro 002	4.1
Claude 3 Haiku	4.0
Qwen2.5 72B Instruct	3.9
Llama 3.1 405B	3.8
DeepSeek-V2.5	3.8
GPT-4o	3.7