Skip to content

Benchmark

MathArena IMO 2025

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

Models measured
6
Scores recorded
7
Sources
1
IQ weight
0.55

Score distribution

20.1%
Mean
16.7%
Median
10.3%
Std dev
11.9%
10th percentile
31.6%
90th percentile
6.8%
Min
6.8%38.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.55 for chat models, measured over 6 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model
1GPT 5high· MathArena data38.1%
2Gemini 2.5 Pro· MathArena data31.6%
3Grok 4Specific Prompt· MathArena data21.4%
4OpenAI o3high· MathArena data16.7%
5o4 Minihigh· MathArena data14.3%
6Grok 4· MathArena data11.9%
7DeepSeek R1· MathArena data6.8%

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)