Skip to content

Benchmark

Creative Writing v3

Models measured
135
Scores recorded
138
Sources
2
IQ weight
1.57

Score distribution

1,302.9
Mean
1,436.1
Median
572.4
Std dev
450.3
10th percentile
1,914.9
90th percentile
0.803
Min
0.8032,176.9
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 135 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Grok 4.20beta· EQ-Bench boards1,573.5
52DeepSeek V4 Flash· EQ-Bench boards1,559.4
53Gemini 3.5 Flash Lite· EQ-Bench boards1,559.3
54DeepSeek V4 Pro· EQ-Bench boards1,552.6
55DeepSeek V4.1 Flash· EQ-Bench boards1,539.9
56Gemini 3 Pro Preview· EQ-Bench boards1,524.3
57DeepSeek V3.2· EQ-Bench boards1,514.9
58ChatGPT-4o Latest· EQ-Bench boards1,500.9
59DeepSeek R1· EQ-Bench boards1,500
60MiMo V2.5 Pro· EQ-Bench boards1,493.1
61Inkling Small· EQ-Bench boards1,491.4
62Gemini 3.1 Pro Preview· EQ-Bench boards1,491.2
63Claude Sonnet 4· EQ-Bench boards1,482.6
64Qwen3.5 397B A17B· EQ-Bench boards1,478.3
65Mistral Medium 3.1· EQ-Bench boards1,475.5
66DeepSeek V3· EQ-Bench boards1,472.2
67Claude 3.5 Sonnet· EQ-Bench boards1,450.9
68DeepSeek V4 Flash 0731· EQ-Bench boards1,441.3
69DeepSeek V3.1· EQ-Bench boards1,436.1
70Gemini 2.5 Pro· EQ-Bench boards1,421.6
71Hermes 4 405B· EQ-Bench boards1,421.1
72GPT 4.1· EQ-Bench boards1,420.2
73GLM 4.7· EQ-Bench boards1,413.6
74Mistral Large 3 675b Instruct· EQ-Bench boards1,412.6
75Claude Sonnet 3.7· EQ-Bench boards1,412.1
76GLM 4.6· EQ-Bench boards1,411.5
77Gemini 2.5 Pro· EQ-Bench boards1,396.4
78Mistral Small Creative· EQ-Bench boards1,369.8
79Gemma 4 31B IT· EQ-Bench boards1,368.6
80Qwen 3 235B A22B· EQ-Bench boards1,366.3
81MiniMax M2.5· EQ-Bench boards1,361.6
82GLM 4.5· EQ-Bench boards1,344
83Grok 4.1 Fast· EQ-Bench boards1,327.9
84GPT-5 mini· EQ-Bench boards1,313.6
85Gemma 4 26B A4B· EQ-Bench boards1,305.1
86Gemma 4 12B It· EQ-Bench boards1,289.5
87Nemotron 3.5 Lightning 30B A3B· EQ-Bench boards1,280.9
88DeepSeek V3.2 Speciale· EQ-Bench boards1,276.9
89Gemma 3 27B IT· EQ-Bench boards1,266.4
90GPT-4.5 Preview· EQ-Bench boards1,258.8
91QwQ 32B· EQ-Bench boards1,258.1
92Mistral Small 3.2 24B Instruct· EQ-Bench boards1,256.2
93Reka Flash 3· EQ-Bench boards1,228.7
94Grok 3beta· EQ-Bench boards1,186.3
95GPT-4.1 mini· EQ-Bench boards1,145.4
96C4AI Command A 03 2025· EQ-Bench boards1,144.7
97Claude Haiku 3.5· EQ-Bench boards1,143.8
98GLM 4.7 Flash· EQ-Bench boards1,138.6
99Gemini 2.5 Flash Preview· EQ-Bench boards1,134.2
100Gemma 3 12B IT· EQ-Bench boards1,131.2

Showing 51–100 of 138 results.

Data credits: EQ-Bench (Samuel J. Paech) — cite paech2023eqbench · Data by LLM Stats