Atlas

Benchmarks

← All benchmarks

EQ-Bench 3: Social IQ

Chat & Writing · 2025-04-28

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

Top models (higher is better)

ModelScore
Opus 4.78.2
Opus 4.88.2
Hivemind-32B-Preview7.8
Sonnet 4.67.7
Claude Fable 57.5
GLM-5.27.5
GPT-5.47.5
GPT-5.57.5
Opus 4.67.4
Gemini 3 Pro Preview7.4
DeepSeek-V4-Pro7.3
Gemma 4 31B IT7.3
GPT-5.17.3
GLM-57.2
GPT-5.27.2
Opus 4.57.1
Gemini 2.5 Pro Preview 06-057.1
Gemini 3.1 Pro Preview6.9
Gemma 4 26B A4B IT6.9
Horizon Alpha6.9
GPT-5 Chat (2025-08-07)6.8
Kimi K2 Instruct6.8
o36.7
ChatGPT-4o Latest observed 2025-03-276.6
DeepSeek-V4-Flash6.6
Gemini 2.5 Pro Preview 05-066.5
Gemini 2.5 Pro Preview 03-256.5
Gemma 4 12B IT6.5
Hermes 4 405B6.5
ChatGPT-4o (2025-04-25 update)6.3
GLM-5.16.3
GLM-4.76.2
Qwen3.5 397B A17B6.2
Kimi K2.56.1
Kimi K2.66.1
Opus 46.0
GPT-4.16.0
Qwen3-235B-A22B6.0
GLM-4.55.9
GPT-5.3 Instant5.9
Loading Atlas data…