Skip to content

Benchmark

MineBench

People compare model answers and choose which they prefer.

Models measured
80
Scores recorded
80
Sources
1
IQ weight
2.01

Score distribution

1,512.8
Mean
1,543
Median
399.5
Std dev
1,014
10th percentile
2,067
90th percentile
544
Min
5442,169
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.01 for chat models, measured over 80 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1GPT-6 Astra· MineBench data2,169
2GPT-6.2 Astra· MineBench data2,152
3Gemini 4 Carbon· MineBench data2,139
4Gemini 4 Argon· MineBench data2,127
5GPT-6.2 Sol· MineBench data2,119
6GPT-6.1 Sol· MineBench data2,116
7Gemini 4 Neon· MineBench data2,091
8Claude Opus 5.5· MineBench data2,079
9Claude Fable 5.2· MineBench data2,067
10GPT-6 Sol· MineBench data2,015
11Gemini 4 Hydrogen· MineBench data1,974
12Claude Opus 5· MineBench data1,959
13Claude Sonnet 5.5· MineBench data1,949
14GPT-5.6 Sol· MineBench data1,946
15Kimi K3.1 Pro· MineBench data1,943
16Claude Fable 5.1· MineBench data1,883
17GPT-5.5 Pro· MineBench data1,878
18Claude Fable 5· MineBench data1,827
19Muse Wonder· MineBench data1,823
20GPT 5.5· MineBench data1,804
21Grok 4.7· MineBench data1,798
22DeepSeek V4.1 Flash· MineBench data1,780
23Grok 4.6· MineBench data1,769
24Gemini 3.8 Flash· MineBench data1,766
25GPT-5.4 Pro· MineBench data1,754
26Claude Opus 4.8· MineBench data1,746
27GLM 5.3 Flash· MineBench data1,739
28Gemini 3.7 Flash· MineBench data1,725
29GLM 5.3· MineBench data1,713
30Muse Spark 1.3· MineBench data1,707
31GPT-6 Luna· MineBench data1,706
32Kimi K3· MineBench data1,676
33GPT-5.6 Luna· MineBench data1,664
34Gemini 3.1 Pro Preview· MineBench data1,639
35GPT 5.4· MineBench data1,629
36Gemini 3.5 Flash· MineBench data1,614
37Claude Sonnet 5· MineBench data1,589
38Gemini 3.6 Flash· MineBench data1,581
39Claude Opus 4.7· MineBench data1,572
40GPT-5.3 Codex· MineBench data1,543
41Claude Haiku 5.5· MineBench data1,539
42Muse Spark 1.2· MineBench data1,529
43GLM 5.2· MineBench data1,524
44GPT-5.2 Pro· MineBench data1,519
45Grok 4.5· MineBench data1,516
46GLM 5.1· MineBench data1,480
47DeepSeek V4 Flash 0731· MineBench data1,459
48Claude Opus 4.6· MineBench data1,452
49GPT 5.2· MineBench data1,436
50DeepSeek V4 Pro· MineBench data1,421

Showing 1–50 of 80 results.

Data credits: MineBench, MIT (licence)