Atlas

Benchmarks

← All benchmarks

Kangaroo 2025 3-4

Math · 2025-10-20

This MathArena track evaluates models on the Kangaroo 2025 3-4 problem set. Scores report the percentage of problems answered correctly.

Top models (higher is better)

ModelScore
GPT-5.589.6
GPT-5.483.3
Kimi K377.1
Gemini 3.1 Pro Preview76.0
GPT-5.274.0
Gemini 3.5 Flash72.9
Gemini 3.6 Flash68.8
Qwen3.5 397B A17B68.8
Gemini 3 Flash Preview66.7
Gemini 3 Pro Preview66.7
GPT-5 Mini66.7
GPT-5.165.6
Kimi K2.565.6
Gemini 2.5 Pro64.6
Sonnet 4.562.5
Opus 4.860.4
GPT-560.4
Qwen3-VL-235B-A22B-Instruct58.3
Grok 452.1
Opus 4.650.0
GLM-4.5V45.8
Grok 4.1 Fast39.6
Grok 4 Fast32.3
Loading Atlas data…