MathArena - Kangaroo 2025 Levels 3-4: leaderboard
Math Kangaroo 2025 problems for grades 3-4, covering basic arithmetic, simple geometry, and logical reasoning.
Metric: Accuracy (%). Source: matharena.ai. Saturation forecast: Around January 2027. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (xHigh) | 89.58 | #31 (GPT-5.5) |
| 2 | GPT-5.4 (xHigh) | 83.33 | #79 (GPT-5.4) |
| 3 | Kimi K3 (Thinking) | 77.08 | #30 (Kimi K3) |
| 4 | Gemini 3.1 Pro (Preview) | 76.04 | #58 |
| 5 | GPT-5.2 (High) | 73.96 | #110 (GPT-5.2) |
| 6 | Gemini 3.5 Flash | 72.92 | #60 |
| 7 | Qwen 3.5 397B A17B | 68.75 | #146 |
| 8 | Gemini 3.6 Flash | 68.75 | #51 |
| 9 | Gemini 3 Flash | 66.67 | #99 |
| 10 | Gemini 3 Pro (Preview) | 66.67 | #66 |
| 11 | GPT-5 Mini (High) | 66.67 | #186 (GPT-5 Mini) |
| 12 | Kimi K2.5 (Thinking) | 65.62 | #144 (Kimi K2.5) |
| 13 | GPT-5.1 (High) | 65.62 | #135 (GPT-5.1) |
| 14 | Gemini 2.5 Pro | 64.58 | #152 |
| 15 | Claude Sonnet 4.5 (Thinking) | 62.5 | #142 (Claude Sonnet 4.5) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=matharena-kangaroo-2025-levels-3-4 · How It Works · Data refreshed daily, snapshot 2026-10-20.