MathArena - Kangaroo 2025 Levels 1-2 — leaderboard

Metric: Accuracy (%). Source: matharena.ai. 21 models tracked.

Top models

#ModelScore
1Human Expert100
2GPT-5.5 (xHigh)95.83
3GPT-5.4 (xHigh)94.79
4Gemini 3.5 Flash89.58
5Gemini 3 Flash87.5
6Gemini 3.1 Pro (Preview)86.46
7Qwen 3.5 397B A17B81.25
8GPT-5.2 (High)80.21
9Gemini 3 Pro (Preview)76.04
10Kimi K2.5 (Thinking)76.04
11Median Human75
12Claude Opus 4.8 (Max)70.83
13GPT-5 (High)68.75
14GPT-5.1 (High)65.62
15GLM-4.5V65.62

Interactive version: theaggregate.ai/benchmark?slug=matharena-kangaroo-2025-levels-1-2 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.