Name: EQ-Bench 3: Moralising
Published: 2025-04-28

EQ-Bench 3: Moralising

Chat & Writing · 2025-04-28

Within EQ-Bench 3's emotional-intelligence scenarios, this subscore rates the moralising dimension of model responses on a 0–10 scale. Lower values are better.

Top models (lower is better)

Model	Score
Gemini 2.5 Flash Preview 05-20	2.5
Gemini 2.5 Pro Preview 03-25	2.6
GPT-5.3 Instant	2.6
GPT-5 Chat (2025-08-07)	2.7
Sonnet 4.6	2.9
Gemini 2.5 Pro Preview 05-06	2.9
Gemma 2 9B IT	3.0
Grok 3	3.0
Horizon Alpha	3.1
Llama 4 Maverick Instruct	3.1
o3	3.2
ChatGPT-4o (2025-04-25 update)	3.3
Opus 4.7	3.4
Opus 4.8	3.4
Gemini 2.5 Pro Preview 06-05	3.4
GPT-5.1	3.4
Llama 3.1 405B Instruct	3.4
GPT-4.1 Mini	3.5
Hivemind-32B-Preview	3.5
Llama 4 Scout Instruct	3.5
GPT-5.2	3.6
Opus 4.5	3.8
Grok 4	3.8
Claude Fable 5	3.9
Gemini 2.0 Flash 001	3.9
GPT-4.1	3.9
GPT-5.4	3.9
O4 Mini	3.9
Qwen2.5 72B Instruct	3.9
ChatGPT-4o Latest observed 2025-03-27	4.0
GLM-5	4.0
gpt-oss-120b	4.0
Grok 4.1 Fast	4.0
DeepSeek-R1	4.1
DeepSeek-V3-0324	4.1
Grok 3 Mini	4.1
gpt-oss-20b	4.2
Kimi K2 Instruct	4.2
Llama 3.2 1B Instruct	4.2
DeepSeek-V4-Flash	4.3