Skip to content

Benchmark

EQ-Bench 3

Vision benchmarks measure how well models understand images.

Models measured
77
Scores recorded
77
Sources
1
IQ weight
1.68

Score distribution

1,260.7
Mean
1,272.6
Median
331.9
Std dev
864.8
10th percentile
1,552.8
90th percentile
200
Min
2002,096.4
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.68 for chat models, measured over 77 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 Fable 5· EQ-Bench boards2,096.4
2Claude Opus 4.8· EQ-Bench boards2,057.8
3Claude Opus 4.7· EQ-Bench boards1,914
4Claude Opus 4.6· EQ-Bench boards1,712.4
5Claude Sonnet 4.6· EQ-Bench boards1,709.8
6GLM 5.2· EQ-Bench boards1,563.2
7DeepSeek V4 Pro· EQ-Bench boards1,555.1
8GPT 5.5· EQ-Bench boards1,553.9
9Gemini 3 Pro Preview· EQ-Bench boards1,552.8
10Kimi K2.6· EQ-Bench boards1,548.3
11Kimi K2 Instruct· EQ-Bench boards1,548.1
12GPT 5.2· EQ-Bench boards1,544.1
13GLM 5.1· EQ-Bench boards1,538.6
14GPT 5.4· EQ-Bench boards1,536.7
15Claude Sonnet 4.5· EQ-Bench boards1,534.6
16Kimi K2.5· EQ-Bench boards1,530.2
17Gemini 3.1 Pro Preview· EQ-Bench boards1,527.2
18GPT 5.1· EQ-Bench boards1,526.6
19Gemini 2.5 Pro· EQ-Bench boards1,522.9
20Claude Opus 4.5· EQ-Bench boards1,521.9
21GPT-5 Chat· EQ-Bench boards1,510.8
22DeepSeek V4 Flash· EQ-Bench boards1,503.6
23OpenAI o3· EQ-Bench boards1,500
24GLM 5· EQ-Bench boards1,495
25Gemma 4 31B IT· EQ-Bench boards1,463.5
26GLM 4.7· EQ-Bench boards1,462.6
27Gemma 4 26B A4B· EQ-Bench boards1,454.6
28Qwen3.5 397B A17B· EQ-Bench boards1,450.6
29GPT-5.3 Chat· EQ-Bench boards1,445.5
30Claude Opus 4· EQ-Bench boards1,440
31Gemini 2.5 Pro Preview 03-25· EQ-Bench boards1,431.2
32ChatGPT-4o Latest· EQ-Bench boards1,427.5
33Gemma 4 12B It· EQ-Bench boards1,415.6
34Hermes 4 405B· EQ-Bench boards1,404.5
35GLM 4.5· EQ-Bench boards1,322
36Claude Sonnet 4· EQ-Bench boards1,304.2
37DeepSeek R1· EQ-Bench boards1,292.4
38o4 Mini· EQ-Bench boards1,278.7
39Gemma 4 E4B-it· EQ-Bench boards1,272.6
40Gemini 2.5 Pro Preview 06-05· EQ-Bench boards1,270.6
41Qwen 3 235B A22B· EQ-Bench boards1,267
42GPT 4.1· EQ-Bench boards1,257.1
43Grok 4.1 Fast· EQ-Bench boards1,202.7
44QwQ 32B· EQ-Bench boards1,184.8
45Gemini 2.5 Flash Preview· EQ-Bench boards1,177.5
46Grok 3beta· EQ-Bench boards1,164.5
47Grok 4· EQ-Bench boards1,154.3
48Claude Sonnet 3.7· EQ-Bench boards1,128.4
49DeepSeek V3· EQ-Bench boards1,123.9
50GLM 4.7 Flash· EQ-Bench boards1,118.4

Showing 1–50 of 77 results.

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