Skip to content

Benchmark

MathArena 12/2025

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

Models measured
19
Scores recorded
20
Sources
1
IQ weight
1.24

Score distribution

43.2%
Mean
41.5%
Median
9.8%
Std dev
32.4%
10th percentile
57.4%
90th percentile
27.9%
Min
27.9%66.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.24 for chat models, measured over 19 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 data66.2%
2GPT 5.4xhigh· MathArena data60.3%
3Claude Opus 4.6high· MathArena data57.4%
4GPT 5.2high· MathArena data52.2%
5Gemini 3 Pro Previewpreview· MathArena data51.1%
6Grok 4.1 Fastreasoning· MathArena data50%
7Gemini 3 Flashhigh· MathArena data43.4%
8Kimi K2.5high· MathArena data41.9%
9Step 3.5 Flash· MathArena data41.9%
10DeepSeek V3.2think· MathArena data41.5%
11Qwen3.5 27B· MathArena data41.2%
12Qwen3.5 35B A3B· MathArena data39.7%
13Qwen3.5 9B· MathArena data39.7%
14GLM 5· MathArena data38.2%
15Qwen3.5 397B A17B· MathArena data38.2%
16Qwen3.5 4B· MathArena data34.2%
17Nemotron 3 Super· MathArena data33.8%
18GPT 5.2low· MathArena data32.4%
19Qwen3 4Bthink· MathArena data32.4%
20Qwen3 30B A3Bthink· MathArena data27.9%

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)