EQ-Bench 4

Emotional intelligence in multi-turn role-play: models handle 120 interpersonal scenarios and judge models compare transcripts pairwise for empathy, insight and social skill.

CategoryGeneral
MetricJudge Elo
SourceEQ-Bench (original)
LicenseNone stated
Benchmark pageeqbench.com
StatusScored
Weight0.8 within General
FitDiscrimination 0.76, noise 0.17

Results

#ModelProviderResultDate
1 Claude Opus 5 Anthropic 1,365 2026-10-24
2 Claude Opus 5.5 Anthropic 1,340 2026-10-24
3 Muse Wonder Meta 1,325 2026-10-24
4 Kimi K3 Moonshot AI 1,319 2026-10-24
5 Claude Fable 5 Anthropic 1,313 2026-10-24
6 GPT-5.5 OpenAI 1,295 2026-10-24
7 Claude Opus 4.7 Anthropic 1,293 2026-10-24
8 GPT-6.1 Sol OpenAI 1,272 2026-10-24
9 Claude Opus 4.8 Anthropic 1,268 2026-10-24
10 GPT-5.4 OpenAI 1,258 2026-10-24
11 Muse Spark 1.1 Meta 1,244 2026-10-24
12 Gemini 4 Argon Google 1,236 2026-10-24
13 GPT-5.6 Sol OpenAI 1,236 2026-10-24
14 Claude Sonnet 5 Anthropic 1,221 2026-10-24
15 GPT-5.6 Terra OpenAI 1,220 2026-10-24
16 Inkling Thinking Machines 1,212 2026-10-24
17 Claude Opus 4.6 Anthropic 1,210 2026-10-24
18 GLM-5.2 Zhipu AI 1,208 2026-10-24
19 MiMo-V2.5-Pro Xiaomi 1,198 2026-10-24
20 Claude Sonnet 4.6 Anthropic 1,197 2026-10-24
21 Kimi K2.6 Moonshot AI 1,191 2026-10-24
22 DeepSeek V4 Pro DeepSeek 1,156 2026-10-24
23 GPT-5.6 Luna OpenAI 1,146 2026-10-24
24 MiniMax-M3 MiniMax 1,140 2026-10-24
25 Gemini 3.1 Pro Preview Google 1,132 2026-10-24
26 Gemma 4 31B IT Google 1,110 2026-10-24
27 Qwen3.7 Max Alibaba 1,100 2026-10-24
28 Gemini 3.5 Flash Google 1,077 2026-10-24
29 Grok 4.3 xAI 1,065 2026-10-24
30 Claude Haiku 4.5 Anthropic 1,053 2026-10-24
31 Qwen3.6 27B Alibaba 1,015 2026-10-24
32 Mistral Medium 3.5 Mistral 983 2026-10-24