Skip to content

Benchmark

Longform Creative Writing

Scores are percentages.

Scores from EQ-Bench boards.
Models measured
124
Scores recorded
124
Sources
1
IQ weight
1.57

Score distribution

59.2%
Mean
60.2%
Median
17.4%
Std dev
32.9%
10th percentile
81.5%
90th percentile
17%
Min
17%86.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.57 for chat models, measured over 124 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Opus 5· EQ-Bench boards86.3%
2Claude Fable 5.1· EQ-Bench boards85.3%
3Claude Opus 5.5· EQ-Bench boards84.6%
4Claude Fable 5· EQ-Bench boards83%
5Grok 4.7· EQ-Bench boards83%
6GPT-6 Astra· EQ-Bench boards82.8%
7GPT-6 Sol· EQ-Bench boards82.8%
8Muse Spark 1.3· EQ-Bench boards82.8%
9Aion 3.5· EQ-Bench boards82.1%
10Claude Opus 4.7· EQ-Bench boards81.8%
11GLM 5.3· EQ-Bench boards81.8%
12GPT-5.6 Sol· EQ-Bench boards81.7%
13Muse Spark 1.2· EQ-Bench boards81.5%
14Gemini 4 Argon· EQ-Bench boards80.9%
15Claude Opus 4.8· EQ-Bench boards80.8%
16Claude Sonnet 4.6· EQ-Bench boards79.9%
17Kimi K3.1 Pro· EQ-Bench boards79.9%
18Kimi K3· EQ-Bench boards79.6%
19Kimi K2.6· EQ-Bench boards78.5%
20Claude Sonnet 5· EQ-Bench boards78.3%
21GPT 5.4· EQ-Bench boards78.3%
22GPT 5.5· EQ-Bench boards78.2%
23GPT-5.6 Terra· EQ-Bench boards78%
24GLM 5.2· EQ-Bench boards77.9%
25Claude Opus 4.6· EQ-Bench boards77.7%
26Gemini 3.8 Flash· EQ-Bench boards76.8%
27DeepSeek V4 Pro· EQ-Bench boards75.6%
28GPT-5.6 Luna· EQ-Bench boards75.5%
29Kimi K2.5· EQ-Bench boards74.9%
30DeepSeek V4.1 Flash· EQ-Bench boards74%
31MiMo V2.5 Pro· EQ-Bench boards73.7%
32GLM 5.1· EQ-Bench boards73.5%
33Claude Opus 4.5· EQ-Bench boards73.1%
34GPT-5.4 mini· EQ-Bench boards73%
35GPT 5· EQ-Bench boards72.8%
36Inkling FP4· EQ-Bench boards72.5%
37Gemini 3.5 Flash· EQ-Bench boards71.8%
38GPT 5.2· EQ-Bench boards71.6%
39Claude Sonnet 4.5· EQ-Bench boards71.4%
40GLM 5· EQ-Bench boards70.9%
41MiMo V2 Pro· EQ-Bench boards70.9%
42Muse Spark 1.1· EQ-Bench boards70.6%
43Claude Opus 4.1· EQ-Bench boards70.5%
44Kimi K2 Thinking· EQ-Bench boards69%
45Gemini 3 Pro Preview· EQ-Bench boards68.8%
46Grok 4.20beta· EQ-Bench boards68.5%
47Qwen3.5 397B A17B· EQ-Bench boards68.3%
48Gemini 3.1 Pro Preview· EQ-Bench boards68.2%
49Kimi K2 Instruct· EQ-Bench boards68.1%
50DeepSeek V3.2· EQ-Bench boards66.8%

Showing 1–50 of 124 results.

Data credits: EQ-Bench (Samuel J. Paech) — cite paech2023eqbench