Atlas

Benchmarks

← All benchmarks

EQ-Bench 4: Bond and Rapport

Chat & Writing · 2026-07-23

EQ-Bench 4's independently fitted bond-and-rapport ability rating measures how well a model establishes trust and emotional safety. The source normalizes its per-dimension soft Bradley–Terry rating to a 1–10 scale.

Top models (higher is better)

ModelScore
Claude Opus 510.0
Kimi K38.9
Claude Fable 58.8
GPT-5.58.8
Opus 4.78.5
Muse Spark 1.17.8
GPT-5.47.7
Opus 4.87.5
GPT-5.6 Terra6.7
GPT-5.6 Sol6.7
Inkling6.5
Sonnet 56.4
GLM-5.26.2
MiMo-V2.5-Pro6.0
Opus 4.66.0
Sonnet 4.65.6
DeepSeek-V4-Pro5.6
Gemini 3.1 Pro Preview5.5
Kimi K2.65.4
Gemma 4 31B IT4.5
MiniMax M34.4
GPT-5.6 Luna4.4
Qwen3.7-Max4.2
Gemini 3.5 Flash3.5
Grok 4.32.9
Haiku 4.51.6
Qwen3.6 27B1.5
Mistral Medium 3.51.0
Loading Atlas data…