Skip to content

Benchmark

MathArena Project Euler

Math benchmarks measure how well models solve maths problems. Scores are percentages.

Models measured
17
Scores recorded
17
Sources
1
IQ weight
1.38

Score distribution

62.6%
Mean
61.5%
Median
17.6%
Std dev
46.1%
10th percentile
87.7%
90th percentile
26.4%
Min
26.4%89%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.38 for chat models, measured over 17 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model
1Gemini 3.1 Pro Previewhigh· MathArena data89%
2GPT 5.4xhigh· MathArena data89%
3Claude Opus 4.6high· MathArena data87.7%
4Gemini 3.5 Flashhigh· MathArena data82%
5GPT 5.2high· MathArena data81.6%
6GLM 5.1· MathArena data67.2%
7Gemini 3 Pro Previewpreview· MathArena data62.4%
8Gemini 3 Flashhigh· MathArena data61.8%
9GPT 5.1high· MathArena data61.5%
10Kimi K2.5high· MathArena data59.9%
11GPT 5high· MathArena data56.9%
12DeepSeek V3.2think· MathArena data50%
13Kimi K2 Thinking· MathArena data49.8%
14o4 Minihigh· MathArena data48%
15Grok 4· MathArena data46.1%
16Grok 4.1 Fastreasoning· MathArena data44.9%
17Gemini 2.5 Pro· MathArena data26.4%

Tokens the model generated per task, as measured by the source. Comparable within this benchmark only.

Data credits: MathArena (ETH SRI / INSAIT), MIT — cite dekoninck2026matharena (licence)