Benchmark
MathArena 03/2026
Math benchmarks measure how well models solve maths problems. Scores are percentages.
All math benchmarksScores from MathArena data.
- Models measured
- 18
- Scores recorded
- 19
- Sources
- 1
- IQ weight
- 1.24
Score distribution
- 13.5%
- Mean
- 14.6%
- Median
- 7.6%
- Std dev
- 5.8%
- 10th percentile
- 19.2%
- 90th percentile
- 0%
- Min
This benchmark has a weight of 1.24 for chat models, measured over 18 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | |
|---|---|---|
| 1 | Claude Opus 4.8max· MathArena data | 35.7% |
| 2 | DeepSeek V4 Flashmax· MathArena data | 19.6% |
| 3 | GPT-5.6 Solmax, via Codex· MathArena data | 19.2% |
| 4 | GPT 5.4xhigh· MathArena data | 17.1% |
| 5 | GPT 5.5xhigh· MathArena data | 17.1% |
| 6 | leanstral-1-5· MathArena data | 17.1% |
| 7 | Gemini 3.5 Flashhigh· MathArena data | 16.1% |
| 8 | Kimi K2.6high· MathArena data | 15.6% |
| 9 | DeepSeek V4 Promax· MathArena data | 15.2% |
| 10 | Claude Opus 4.7high· MathArena data | 14.6% |
| 11 | Gemini 3.1 Pro Previewhigh· MathArena data | 14.6% |
| 12 | GLM 5· MathArena data | 12.1% |
| 13 | GLM 5.2max· MathArena data | 12.1% |
| 14 | Step 3.7 Flashhigh· MathArena data | 10.7% |
| 15 | Qwen3.5 2B· MathArena data | 6.7% |
| 16 | Claude Opus 4.6high· MathArena data | 5.8% |
| 17 | Claude Opus 4.7xhigh· MathArena data | 5.8% |
| 18 | GLM 5.1· MathArena data | 2.4% |
| 19 | Step 3.5 Flash· MathArena data | 0% |
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)