Skip to content

Benchmark

MathArena USAMO 2026

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

Models measured
9
Scores recorded
9
Sources
1
IQ weight
0.99

Score distribution

60.3%
Mean
51.2%
Median
22.5%
Std dev
36.3%
10th percentile
95.2%
90th percentile
35.1%
Min
35.1%98.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.99 for chat models, measured over 9 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 5.5xhigh· MathArena data98.2%
2GPT 5.4xhigh· MathArena data95.2%
3Gemini 3.1 Pro Previewhigh· MathArena data74.4%
4DeepSeek V4 Promax· MathArena data60.7%
5Kimi K2.6high· MathArena data51.2%
6Claude Opus 4.6high· MathArena data47%
7Step 3.5 Flash· MathArena data44.6%
8Qwen3.5 397B A17B· MathArena data36.3%
9GLM 5· MathArena data35.1%

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)